samfunnsokonomisk-analyse.no
f
01.10.2018
Evaluation of the regionally differentiated social security contributions in Norway Report 26-2018
Report no. 26-2018 from Samfunnsøkonomisk analyse AS ISBN-number:
978-82-8395-017-5
Principal:
Ministries of Finance and of Local Government and Modernisation
Cover photo:
iStock
Accessibility:
Public
Date of completion:
1 October 2018
Authors:
Andreas Benedictow, Emil Cappelen Bjøru, Fernanda Winger Eggen, Vegard Salte Flatval, Marthe Norberg-Schulz, Marina Rybalka, Rolf Røtnes, Arne Stokka, Maja Tofteng and Lars Vik
Samfunnsøkonomisk analyse AS Borggata 2B N-0650 Oslo VAT number: NO 911 737 752
[email protected]
Preface
The Norwegian Ministries of Finance and of Local Government and Modernisation have commissioned Samfunnsøkonomisk analyse AS to evaluate the Norwegian scheme of regionally differentiated social security contributions. The evaluation has been completed in accordance with the European Commission’s Common methodology for State aid evaluations. As project manager, I would like to acknowledge the substantial input from my colleagues, Marina Rybalka, Marthe Norberg-Schulz, Emil Cappelen Bjøru, Fernanda Winger Eggen, Vegard Salte Flatval, Maja Tofteng and Rolf Røtnes at Samfunnsøkonomisk analyse AS. We would also like to thank Ragnar Nymoen for contributions to the theoretical and empirical frameworks, Michael Lechner as our project advisor and Michael Spjelkavik Mark and Roger Bjørnstad as project participators at an early stage. SINTEF Technology and Society contributed to the evaluation with an analysis of the scheme’s ripple effects. We would like to thank Arne Stokka and Lars H. Vik for a fruitful collaboration. We thank the reference group members from the Norwegian Ministries of Finance, Local Government and Modernisation and Trade, Industry and Fisheries, and the EFTA Surveillance Authority for interesting discussions and useful feedback. We especially want to thank Jan Oosterhaven from the University of Groningen and Frode Steen from NHH Norwegian School of Economics for helpful comments and suggestions as external referees for a previous draft. Samfunnsøkonomisk analyse AS is responsible for the entire content of this report. Oslo, 1 October 2018 Andreas Benedictow Project manager Samfunnsøkonomisk analyse AS
IV
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
Abstract
Social security contributions through employer-paid
makes an important contribution to maintaining ac-
payroll tax have been regionally differentiated in
tivity and employment, especially where the rates
Norway since 1975. The objective of regionally dif-
are low or zero. Thus, we cannot conclude from our
ferentiated social security contributions (RDSSC) is
moderate estimated effects that the impact of the
to reduce or prevent depopulation in the most
scheme is low. The estimates should rather be con-
sparsely populated regions of Norway by stimulat-
sidered conservative.
ing employment through reduced employment costs. The scheme is the most comprehensive re-
We find that RDSSC does enhance beneficiaries’
gional policy measure in Norway.
competitiveness domestically, which is the intention of the scheme. Most firms receiving aid from
Using detailed micro data on Norwegian firms and
RDSSC offer services locally, clearly limiting the po-
employees and state of the art empirical methods to
tential impact on international competition and
study changes in the scheme, we find evidence that
trade. The share of export-oriented firms is not sig-
most of the tax incidence resides with the employ-
nificantly higher in the zones with reduced rates.
ers, implying that a change in payroll tax has rela-
Furthermore, the exporting firms in the zones with
tively limited effects on wages. We find that employ-
reduced rates tends to be capital-intensive, thus
ment increases directly because of reduced wage
gaining relatively little from a tax scheme that re-
costs, allowing firms to reduce product prices to in-
duces the relative cost of labour. We also argue that
crease production and gain market share. The
the scope of import competition is limited by a high
scheme also contributes indirectly to increased em-
level of specialisation and low intra industry trade.
ployment by shifting some of the tax reduction on to
The evaluation also finds that the vast majority of
workers through higher wages, thereby increasing
exporting firms receive support that is under the
household demand for locally produced goods and
threshold for de minimis aid and is thus not defined
services. In addition to positive effects on employ-
as anti-competitive state aid according to EU rules.
ment in existing firms, a descriptive analysis indi-
We conclude that there is little evidence of RDSSC
cates that employment also increases through the
having a distortive impact on competition and trade
establishment of new firms.
to an extent contrary to the intent of the EEA agreement.
Overall, our estimates indicate moderate employment effects. However, it is reasonable to assume
We put forward a clear recommendation of continu-
that the effects of changes in payroll tax are not lin-
ing the scheme, although we also suggest consid-
ear. A small change could be expected to have a
ering the possibility of allowing municipalities to
small or no effect, because the risk and costs asso-
choose between RDSSC and receiving the same
ciated with reallocating resources reduce firms’ in-
amount of support but in the form of separate free
centives to change their behaviour. However, if pay-
income. We suggest that such an option be limited
roll tax had increased from the lowest to the highest
to municipalities in the zones with tax rates close to
tax rate, i.e. from 0 to 14.1 per cent, we would ex-
the highest level.
pect substantial effects. This is supported by our calculations, which show that such a jump would increase the share of firms with negative operating profits considerably. This suggests that RDSSC
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
V
Executive summary
that RDSSC contributes to reducing or preventing
1) Does RDSSC have a well-defined objective of common interest?
depopulation in the regions eligible for the scheme.
RDSSC is the single most comprehensive regional
We submit a clear recommendation that the scheme
policy measure in Norway and has been part of a
be continued, although we also suggest considering
broad regional policy since its introduction in 1975.
the possibility of allowing municipalities to choose
The policy finds legitimacy through broad popular
between RDSSC and receive the same amount of
and political support. The objective of RDSSC is to
support but in the form of separate free income. We
reduce or prevent depopulation in the most sparsely
suggest that such an option be limited to municipal-
populated regions in Norway by stimulating employ-
ities in the zones with tax rates close to the highest
ment.
Based on an extensive empirical review, we find
level. The Norwegian economy is characterised by low laThe Norwegian system of social security contribu-
bour mobility and a national collective wage bar-
tions is divided into seven different zones with rates
gaining system, the latter leading to a relatively high
varying from 14.1 pct. in central areas to 0 pct. in
degree of wage equalisation for equal work across
the northern most part of the country. This report
geographical regions. As a result, wages do not per-
evaluates the scheme of regionally differentiated
fectly reflect the scarcity of production factors. In re-
social security contributions (RDSSC).
mote areas with small labour markets and/or a onesided industrial base, this could typically result in
The Norwegian authorities notified the current
higher wages and lower employment than would
scheme for the period 1 July 2014 to 31 December
have resulted from perfect competition. Under such
2020 to the EFTA Surveillance Authority (ESA) in
circumstances, subsidising wages to offset the gap
2014. As part of the notification, the Norwegian au-
between tariff and market wages could offset rela-
thorities undertook to evaluate the scheme, in ac-
tively high labour costs in rural areas.
cordance with ESA’s Regional Aid Guidelines (RAG). The Ministry of Finance and the Ministry of
When RDSSC was introduced, the differentiation of
Local Government and Modernisation commis-
tax rates was justified by a reduction in employment
sioned Samfunnsøkonomisk analyse AS and SIN-
in primary industries in rural areas. In combination
TEF Technology and Society to conduct the evalu-
with low labour mobility across regions and nation-
ation.
ally determined wages, this could create “hidden” unemployment. This may still be the case, but the
In line with the objective of the evaluation as stated
arguments for stimulating rural employment have
by the Ministry of Finance and the European Com-
changed over the years. Today the main argument
mission Staff Working Document, Common meth-
is the importance of stimulating rural employment to
odology for State aid evaluations, the evaluation
avoid depopulation, and RDSSC is justified as com-
has tested and analysed whether RDSSC 1) has a
pensation for lower productivity in rural areas due
well-defined objective of common interest, 2) is de-
primarily to poorer infrastructure and lack of econo-
signed to achieve the objective of common interest,
mies of scale.
3) is appropriate and correctly proportioned for
VI
achieving these targets and 4) has a distortive im-
Whether and to what extent the objective of prevent-
pact on competition and trade.
ing or reducing depopulation is achieved through
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
RDSSC therefore depends on a positive relation-
least populated regions of Norway will contribute to
ship between employment and population.
reducing, or preventing, depopulation in the eligible regions.
The reasoning behind RDSSC is that the scheme will counteract discrepancies between market earn-
Employment may be increased directly by RDSSC
ings and actual wages due to centralised wage for-
reducing labour costs, allowing firms to reduce
mation and weak mobility in rural areas. However,
product prices to increase production and gain mar-
the objective of RDSSC is politically determined, in
ket share. RDSSC may also contribute to increased
contrast to most industrial policy schemes, which
employment indirectly, if part of a tax reduction is
are usually justified as mitigating market failures.
shifted to workers through higher wages, which in
This does not undermine the legitimacy of RDSSC,
turn leads to increased demand for goods and ser-
although the academic justification is weaker.
vices, activity and employment in the local economy. It is the direct effect that explicitly justifies the
Overall, the objective of the scheme is clear and
choice of RDSSC as a policy instrument. The indi-
easily understood and is sought accomplished
rect effect might as easily, and maybe more effec-
through theoretically convincing means. Further, the
tively, be achieved by other means, addressing
scheme has broad and long-standing political sup-
worker or household income directly.
port. We therefore conclude that RDSSC addresses a well-defined objective of common interest.
In our empirical analyses, we use detailed micro data on Norwegian firms and employees and state
2) Is RDSSC designed to achieve the objective of common interest? The scheme is designed to stimulated employment by offsetting employment costs. While employment is directly influenced by the payroll tax rate through labour costs, multiple factors can affect the population in the eligible regions: various demographic factors such as birth and death rates, migration, civil status, etc., making it hard to estimate the effect on population directly. Therefore, our approach to study effects of RDSSC on population in the most sparsely populated regions in Norway is to estimate the effect on employment. A wide range of studies have analysed the interdependent processes of population and employment growth. They mainly suggest that population and employment are the subject of a dynamic adjustment process and are determined jointly. Despite varying evidence from the literature, aggregated studies suggest that stimulating job creation in the
of the art econometric methods to study the effects of changes in the payroll tax rate on wages, employment, value added and capital. We primarily focus on identifying the impact of differentiated payroll tax rates on wages and employment. Our chosen econometric approaches are mostly in line with previous studies of Norwegian payroll tax. In addition, we apply a regression kink design (RKD), which, at least to our knowledge, has not been used to study the effects of the Norwegian payroll tax scheme before. Our main identification strategy is to use variation induced by different changes in the scheme, socalled exogenous shocks. There have been several changes in the scheme since the introduction of differentiated payroll tax rates in 1975. We exploit the three reforms of the scheme that took place in the period 2000-2007: (i) we use difference-in-differences to study effects of a lower tax rate for firms in municipalities that changed tax zone in 2000; (ii) we use both difference-in-differences and a regression
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
VII
kink design to evaluate effects of increased payroll
costs following changes in the payroll tax affects
taxes in the period 2004-2006; (iii) we exploit all var-
firms’ demand for labour. In total, we find that
iation in the tax rates following the reform in 2004
changing the payroll tax rate with one percentage
and its reversion in 2007 to estimate long-run ef-
point changes employment in the affected firms with
fects on the demand for labour using a GMM esti-
approximately one per cent.
mator. In addition to affect employment in existing firms, Based on international studies of comparable
changes in the payroll tax may affect firm entries
schemes (Saez, Matsaganis and Tsakloglou 2012,
and exits. In a descriptive analysis, we find that be-
Saez, Schoefer and Seim 2017) and previous stud-
tween 2000 and 2002 the share of new firms was
ies of the Norwegian scheme (Johansen and Klette
larger in the treatment group, suggesting that the
1997, Gavrilova, et al. 2015, Stokke 2016, Ku,
tax rate reduction in 2000 had a positive impact on
Schönberg and Schreiner 2018) we expect to find
firm entry. An econometric study of a pay roll tax re-
effects on both workers’ wages and employment.
duction in northern Sweden finds evidence of increased employment through establishments of
As expected and in line we Stokke (2016), we find
new firms (Bennmarker, Mellander and Öckert
positive effects on employment among firms that
2009).
faced a lower tax rate after the change in 2000, both on the extensive and intensive margin. That is, both
We are unable to estimate effects for tax zones
the number of employees and number of hours
where there have been no changes in the payroll tax
worked increased more among firms experiencing a
rate, i.e. Zone 1 and 5. Municipalities in Zone 1
reduction in the payroll tax rate than what would
faces the general payroll tax rate of 14.1 per cent
have been the case without the reduction. We also
and is outside the geographical scope of RDSSC.
find that some of the tax reduction was shifted onto
However, Zone 5 face a zero per cent payroll tax
workers’ wages; about 24 per cent of the tax inci-
rate and is the region with largest differentiation.
dence resides with the employees.
Looking at the increase in the share of firms with negative operating profits if the scheme was abol-
Applying the regression kink design (RKD) to study
ished, it is apparent that the increase is highest in
the changes in tax rates between 2004 and 2006,
Zone 5.
we find that somewhere between 0.5 and 4 pct. of the tax increase was shifted onto workers in Zone 4
Overall, our results clearly indicate that RDSSC has
and between 4 and 17 pct. in Zone 2. Thus, it seems
the intended effects on the beneficiaries. In other
that firms in different regions react differently to
words, the scheme does reduce or prevent depop-
changes in the payroll tax rate.
ulation in the eligible regions. Therefore, we conclude that the scheme is designed to achieve the
Inspired by Johansen and Klette (1997) and Gav-
objective of common interest.
rilova, et al. (2015), we apply a two-step approach to estimate the long-run effects on labour demand. First, we find little or no effect on workers’ wages, meaning that the tax incidence resides with the employers. Second, we find that changes in wage
VIII
3) Is RDSSC appropriate and correctly proportioned to achieve these targets? We interpret this as an assessment of whether the objective could be achieved in a more effective way
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
by other means. To this end, it is useful to keep in
the same effect. Our assessment is that there is little
mind what would have happened without the
to be gained by reorganising in this way.
scheme and what alternative schemes are available or feasible.
Another alternative might be to increase income support to households directly, as is already done in
First, we find that RDSSC contributes to reducing or
Zone 5, especially in regions where a large share of
preventing depopulation in the eligible regions. It fol-
the tax subsidy is shifted to workers anyway. In-
lows directly from the results discussed above that
creased income support may boost regional settle-
repealing the regional differentiation of the social
ment in two ways. First, through the same income-
security contributions within a tax neutral framework
employment effect as higher wages through
would have resulted in lower employment and set-
RDSSC and second, by making it more attractive to
tlement in the eligible regions and higher employ-
live in the eligible regions. However, income support
ment and settlement in Zone 1.
to households will not result in direct employment effects. It is difficult to imagine that income support
Alternative measures may also achieve similar re-
will be more effective than the RDSSC scheme.
sults. However, RDSSC is, in monetary terms, by far the most important scheme within the portfolio of
Transferring the support to the municipalities di-
rural and regional development policies. Moving all
rectly would enable them to increase the employ-
regional support from RDSSC to other schemes
ment related to their responsibilities, invest in com-
would therefore radically change all of them. This
munal goods in the municipality or boost small mu-
raises a serious question about appropriateness.
nicipal industrial funds where any such exist. Better
Normally there would be a decreasing return on
municipal services or communal goods can be fac-
public schemes. If a scheme increases substantially
tors that help to keep or attract labour. However, this
in size, it is reasonable to assume that “the last mil-
would likely shift employment from the commercial
lion” will have very little effect.
sector to the public sector, which in the long run may weaken the ability of rural regions to develop new
Thus, alternatives to RDSSC should preferably be a
income opportunities. Nevertheless, this could be a
mix of other schemes to boost employment and set-
possible alternative to RDSSC as it is today.
tlement in the eligible regions. For instance, capital and innovation subsidies could be increased in eli-
Our assessment is that a total (revenue neutral)
gible regions to promote employment. Innovation
abolishment of RDSSC would clearly weaken the
Norway and the Research Council have several
possibilities of achieving the stated regional policy
such schemes readily available. Evaluations indi-
objectives. The effects would be particularly large in
cate that such schemes affect employment in a sim-
Zone 5. As part of an ambitious regional policy,
ilar way to RDSSC. However, these schemes are
RDSSC appears to be appropriate in combination
much smaller in scope, and we do not know whether
with other schemes. However, it is interesting to
the effects will persist if they are inflated. This would
consider whether some municipalities may be better
particularly be the case in Zone 5, where abolishing
off with a different mix of policy instruments.
RDSSC would increase social security contributions the most and where alternative schemes would
Data shows limited correlation between the zone di-
have to increase relatively substantially to achieve
vision and the centrality index. We argue that there
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
IX
may be grounds for reassessing at suitable intervals
Moreover, we find that very few of the exporting
the zone division in the light of demographic
beneficiaries receives support above the limit of de
changes, for example.
minimis aid, under which support is not defined as distortive state aid according to the EEA agreement.
Overall, we argue that RDSSC seems appropriate and reasonably proportioned. However, as the
We are not able to quantify the scope of import com-
scheme may be less efficient in certain municipali-
petition due to lack of data. However, we argue that
ties, we suggest considering the possibility of allow-
the fact that the economy is small and specialised
ing them to choose between RDSSC and receiving
in industries where Norway has a comparative ad-
the same amount of support transferred in the form
vantage and the relatively low extent of intra-indus-
of separate free income.
try trade in Norway limits the scope of import competition.
4) Does RDSSC have distortive effects on competition and trade? State aid that limit competition are prohibited by the EEA agreement. However, state aid facilitating the development of economic areas, where such aid
We conclude that there is little evidence of RDSSC having a distortive impact on competition and trade to an extent contrary to the intent of the EEA Agreement.
does not adversely affect trading conditions to an extent contrary to the common interest, may be con-
Recommendations
sidered compatible with the functioning of the EEA
We are not able to test the effect of the scheme in
agreement.
the region where the scope is greatest, i.e. in Finnmark and northern Troms (the Action Zone). This is
X
In line with the scheme’s objective, we find that
due to a lack of variation in the scheme in this area
RDSSC does enhance beneficiaries’ competitive-
during the evaluation period. It is reasonable to as-
ness domestically. The scheme is compensating
sume that the effects of changes are not linear. A
firms in more rural areas for having competitive dis-
small change could be expected to have a small or
advantages through larger distances and ineffi-
zero effect because the risk and costs associated
ciently high wages. However, our analysis indicate
with reallocating resources reduce firms’ incentives
that ripple effects reduce the domestic distortive ef-
to change their behaviour. But if the payroll tax had
fect of RDSSC somewhat, as they generate import
suddenly increased from 0 to 14.1 per cent in the
“leakage” by creating increased demand in sur-
Action Zone, for example, we would expect sub-
rounding zones and Zone 1 in particular.
stantial effects.
Most firms receiving aid from RDSSC are offering
RDSSC is the single largest rural policy scheme in
services locally, which reduces the potential impact
Norway, but data variation within our data period is
on international competition and trade. We also find
limited. This makes it challenging to identify effects.
that the proportion of export-oriented firms is not
Our estimates may not reflect the size of the
significantly higher within the zones with reduced
scheme, but rather the relatively limited adjustments
rates. Furthermore, the exporting firms tend to be
of the scheme within the sample period. In other
capital intensive, thus gaining relatively little from a
words, we cannot conclude from our modest esti-
tax scheme reducing the relative cost of labour.
mated effects that the impact of the scheme is small.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
Our estimates should rather be considered as conservative. This is supported by our analysis showing that an increase in the pay roll tax from 0 to 14.1 per cent in the Action Zone would substantially increase the share of firms with negative operating profits, which implies a potential for corresponding negative effects on employment. We find effects on both wages and employment, indicating that there are direct as well as indirect effects on employment and population in the eligible areas. There is good reason to believe that the overall effect of the scheme is significant, especially in the zones with the lowest payroll tax rate. Furthermore, the scope of distortive effects on competition and trade appears to be limited. Based on an extensive empirical review of RDSSC, we recommend that the scheme be continued. However, since some municipalities are experiencing challenges not covered by RDSSC, we suggest that the relevant ministries consider giving individual municipalities the freedom to choose whether they will carry on with RDSSC, or whether they want the same amount of support transferred in the form of separate free income for the municipality. This could, for example, take the form of a pilot scheme to test the municipalities’ interest, but with the opportunity to revert to the previous arrangement later on.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
XI
Contents
Preface
IV
Abstract
V
Executive summary
VI
1
Introduction
14
1.1
Evaluation of the scheme
14
1.2
Outline of the report
15
2
3
4
5
6
Regional Differentiated Social Security Contributions
16
2.1
Initial rationale of the scheme
16
2.2
Administration of the scheme
19
2.3
Changes in the scheme
22
2.4
Seven different tax zones
23
Theoretical framework
27
3.1
Demand for labour
28
3.2
Supply of labour
29
3.3
Market equilibrium
30
3.4
Effects with centralised wage negotiations
32
3.5
Time and direct versus indirect effects
34
3.6
A stylised model
35
Empirical evidence
37
4.1
Evaluation of the reform in 2000
40
4.2
Evaluation of the reform in 2004
59
4.3
Estimating long-run effects
73
4.4
Analysis of larger shifts in the payroll tax
80
The dynamics of regional population growth
81
5.1
Population growth through employment
81
5.2
Urbanisation, productivity and regional development
82
5.3
The current settlement pattern mainly determined by historical events
83
5.4
Circular causation between population growth and employment growth
84
5.5
Mixed evidence from the literature
86
5.6
Complex results call for complex policy measures
89
Alternative measures
90
6.1
The single largest rural policy measure
90
6.2
Economic theory suggests that RDSSC is the most effective policy measure
92
6.3
Experience reveal pitfalls in measures to change factor prices
94
6.4
Individual measures and municipal funding can also prevent depopulation
99
7
8
9
10
Ripple effects
103
7.1
Ripple effects in the analysis
103
7.2
The PANDA models
105
7.3
Ripple effects in the five payroll tax zones
106
7.4
Ripple effects of the 2000-reform
107
7.5
Value added effects
109
7.6
Ripple effects in eligible and non-eligible industries
111
7.7
Income and consumption effects
111
7.8
Income effects
111
Impact on competition and trade
112
8.1
Impact on domestic competition
113
8.2
Impact on international competition
115
Concluding remarks and recommendations
124
9.1
The objective of the scheme is well defined
124
9.2
RDSSC increases employment in eligible regions
125
9.3
Alternatives to RDSSC are costlier and less appropriate
126
9.4
Small effects on competition and trade
128
9.5
Recommendations
129
References
130
Appendix 1: Data
135
Appendix 2: Models and data in PANDA
136
1
Introduction
Most OECD countries fund social insurance pro-
The payroll tax is differentiated according to the pe-
grams, such as retirement, health, disability, and
riphery of the region (measured by geography, de-
unemployment benefits, with substantial social se-
mography, labour market and income) and is lower
curity contributions on employment earnings (Saez,
in rural than in central areas. Today Norway is di-
Matsaganis and Tsakloglou 2012). Social security
vided into seven different zones with rates varying
contributions (payroll taxes) collect about 26 pct. of
from 14.1 pct. in central areas (Zone 1) to 0 pct. in
total tax revenue on average in OECD countries and
the northern most part of the country (Zone 5).
is the second largest source of tax revenue, after tax on goods and services (OECD 2018).
1.1
Evaluation of the scheme
An employer-paid payroll tax was introduced in Nor-
The scheme with regionally differentiated social se-
way in 1967. In 2016 this tax constituted almost 28
curity contributions (RDSSC) has undergone sev-
pct. of total tax revenues.
eral changes since its introduction, both in terms of eligible regions (municipalities) and the difference in
A key objective of the Norwegian post-war eco-
tax rates. The Norwegian authorities notified the
nomic policy has been to achieve full employment
current scheme for the period 1 July 2014 to 31 De-
for the country as a whole (NOU 1975: 2), and
cember 2020 to EFTA Surveillance Authority (ESA)
preservation of the distinctive features of the Nor-
on 3 June 2014.1 As part of the notification, the Nor-
wegian settlement patterns has been an explicit ob-
wegian authorities committed to evaluate the
jective for Norwegian regional policy since the
scheme, in accordance with ESA’s Regional Aid
1970s (Meld. St. 13 (2012-2013)).
Guidelines (RAG).
Following a discussion of labour subsidies as a re-
The Ministries of Finance and of Local Government
gional policy measure, the payroll tax was regionally
and
differentiated in 1975. The differentiation was in line
funnsøkonomisk analyse AS and SINTEF Technol-
with economic theory, showing that a reduction in
ogy and Society to conduct the evaluation. The
labour taxes would be better suited than capital sub-
main assignment is an ex post evaluation of the
sidies to stimulate regional employment in an econ-
scheme, i.e. identify the causal impact of the
omy with high capital mobility, low labour mobility
scheme on the policy objective and quantify its ef-
and a national collective wage bargaining system.
fects.
Lower marginal labour costs in a selected area can
Specifically, as stated by the terms of reference, the
lead to higher employment in the same area, partly
objective of this evaluation is to (1) assess the im-
through the opportunity to increase production and
pact on job opportunities and employment in the el-
partly because it will be profitable to replace capital
igible regions by differentiated rates in the scheme,
with labour in production, to the extent that it is tech-
and whether and to what extent, the objective of pre-
nically possible (NOU 1975: 2).
venting or reducing depopulation is achieved. The
Modernisation
has
commissioned
Sam-
evaluation should assess the incentive effect of the
1
The scheme was notified by letter 13 March 2014 and the notification was completed, after submitting an updated notification, on 3 June 2014.
14
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
implicit aid following reduced rates and to what ex-
We assess the ripple effects of the scheme, i.e. the
tent it changes the behaviour of firms (and employ-
scheme’s indirect impact in Chapter 7, and in con-
ees), (2) assess the effects on competition and
tinuation of this we discuss potential distortive ef-
trade, (3) assess the effects of the new sector limi-
fects in Chapter 8.
tations, and (4) assess whether the objective could be reached in a more effective and less distortive
We conclude with the main results, their implication
way by other means.
and policy recommendations in Chapter 9.
To answer all tasks in a structured manner, we have organised the evaluation according to the European Commissions’ Common methodology for State aid evaluations (European Comission 2014). This means that the evaluation is structured around describing the objectives of the scheme to be evaluated, assessing the direct impact of the aid on beneficiaries, assessing the indirect impact of the scheme and assessing the proportionality and appropriateness of the scheme.
1.2
Outline of the report
The following chapter presents the background for regionally differentiated social security contributions in Norway and changes in the scheme over time. In Chapter 3 we provide a theoretical framework to illustrate how the scheme is intended to work and empirically testable hypotheses. Both chapters cover the objectives of the scheme. In Chapter 4 we present empirical results on the direct impact of the scheme on beneficiaries (e.g. effect on wages, employment, value added and capital). This is followed up in Chapter 5, where we present existing literature on the links between employment and population. In Chapter 6 provides a summary of other (alternative) measures and discusses the proportionality and appropriateness of RDSSC.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
15
2
Regional Differentiated Social Security Contributions
Social security contributions (through employer-
they would have if wages were adjusted freely ac-
paid payroll tax) have been regionally differentiated
cording to local demand conditions (NOU 1975: 2).
in Norway since 1975. The scheme is the most comprehensive regional policy measure in Norway.
The introduction of a regionally differentiated payroll
Prior to the introduction of the regionally differenti-
tax in Norway was based on a series of theoretical
ated payroll tax, regional policy measures were
studies that discussed the market failures of the re-
mainly targeted at supporting investments (NOU
gional labour markets, including the disparity be-
1975: 2).
tween regional demand for labour and nationally determined wages (Hervik and Rye 2010).3
The objective of the regionally differentiated payroll tax is to reduce or prevent depopulation in the most
Johansen (1965) showed that if the objective is
sparsely populated regions in Norway by stimulating
maximising total income, calculation prices provid-
employment. The scheme is designed to offset em-
ing the optimal solution must be such that common
ployment costs. It is estimated a tax relief of about
(mobile) resources have the same calculation rates
NOK 13.9 billion for the whole scheme in 2018, of
in all regions, while regional (immobile) resources
which tax relief to the private sector account for
generally have different calculation rates across re-
NOK 8.2 billion (Prop. 1 S (2017-2018)).
gions. Considering this, Johansen questioned the subsidisation of common resources, such as capi-
In the following we present some of the arguments
tal, rather than labour, which was assumed less mo-
that was used to justify the implementation of the
bile (or even immobile). He further specified that the
scheme, how it is administered and changes over
actual wage paid to (equal) workers did not have to
time. Further, we present the seven current tax
be different in different regions to satisfy the opti-
zones and some characteristics of these.
mum requirements, but the firms’ calculation cost of labour.
2.1
Initial rationale of the scheme The demand for a production factor (input) normally
In an economy with perfect competition, prices for
depends on the price of the product (output) and the
mobile products and input factors will be equal eve-
relative price ratio between the relevant input factor
rywhere, while prices for immobile products and
and all other factors of production. Thus, when the
production inputs may vary (e.g. due to differences
cost of labour changes due to a reduction in the la-
However, when wages are largely de-
bour tax, firms’ optimal adjustment changes, and in
termined through centralised wage bargaining,
turn the demand for different factors of production
equal wages will arise for the same type of work
and level of production (NOU 1975: 2).4
in
demand).2
throughout the country. Firms may then pay higher wages, and thus have lower employment, than what
2
With perfect competition one often assumes no economies of scale and, thus, that there will be enough suppliers of the different products and input factors. However, with economies of scale the locations of production are likely to differ from locations of consumption and prices will differ depending on unit transportation cost and distances. 3 For a more comprehensive description of the prelude to the scheme and different theoretical perspectives, we refer the reader to NOU 1975: 2 and
16
Hervik and Rye (2010). Both in Norwegian. An English summary of the latter can be found online: “An empirical and theoretical perspective on regional differentiated payroll taxes in Norway”. 4 Mechanisms leading to these changes is elaborated in Chapter 3.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
Given the above one could argue that subsidising
cost (time), which in turn gives access to a wider
capital would also lead to increased demand for la-
range of interesting work opportunities.
However, several studies showed that under
certain assumptions, labour subsidies was preferable to capital subsidies (e.g. Serck-Hanssen (1971)). Serck-Hanssen (1982) argued that the reason one should subsidise labour and not the use of capital, when settlement is the objective, is not that it is impossible to achieve this objective by subsidising capital. Increased settlement (or at least reduced depopulation) could be achieved by other means than labour subsidies. It is only more expensive (or equally expensive) to use capital subsidies to
Figure 2.1 Migration per 1,000 mean population. 1975-2017 50
50
40
40
30
30
20
20
10
10
0
0
1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017
bour.5
achieve an employment target (Serck-Hanssen
Between municipalities
1971, 15). He further points out that how much more
Between counties
expensive it will be, depends on how the opportuni-
Between five regions
ties for production are in the region.
Figure 2.2 Commuting as share of total employment. 2000-2017 2.1.1 Labour mobility
40%
35%
35%
payroll tax, labour was considered immobile be-
30%
30%
tween regions. Studying the migration between mu-
25%
25%
nicipalities, counties and the five regions of Norway,
20%
20%
there are few indications that workers are more mo-
15%
15%
bile today than in 1975 (cf. Figure 2.1). However,
10%
10%
migration per 1,000 mean population between mu-
5%
5%
nicipalities and counties has been higher the last
0%
0%
decade than the average for the period 1975-2017.
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
40%
When implementing the regionally differentiated
Between municipalities
Further, there have been tendencies towards more commuting between municipalities compared to the beginning of the 2000s (cf. Figure 2.2). More commuting between municipalities suggest larger labour market regions. More and higher quality infrastructure enables longer commuting distances at a given
Between counties Between five regions Note: Oslo and Akershus is considered as one county. The five regions are Eastern Norway, Agder-Rogaland, Western Norway, Trøndelag and Northern Norway. Source: Statistics Norway
5
A reduction in the price of another factor of production will increase employment if labour is complementary (in production) to the factor being subsidised.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
17
Such access is a decisive factor when choosing
stayed. Thus, it seems reasonable to claim that em-
where to live. Easy access to work through commut-
ployment is not necessarily a sufficient factor in the
ing increase the number of possible places of resi-
decision about whether to move or not.6
dence. The share of workers commuting between counties and regions are unchanged during the same period.
2.1.2 Regional unemployment When the differentiated payroll tax was introduced
Migration between labour markets are not fully captured by migration between municipalities or counties. However, increased migration between municipalities, together with increased commuting, may indicate that the choice of workplace and residence are two separate decisions within a functional labour market region. People can change workplace without moving. Labour market regions may overlap both municipality and county borders. The same explanation can thus be made for migration between
in 1975, overall unemployment was 2.3 pct. 7 The conception was that full employment was achieved, partly through extensive migration from weakly developed regions to central areas. However, there was a concern that different forms of “hidden” unemployment was present in regions where primary industries had previously been a significant employer, as well as underemployment among specific groups of workers in regions with narrow employment opportunities (NOU 1975: 2).
counties. Oslo and Akershus accounted for more than one third of the domestic migration in 2017 and a significant share of it is migration between these two counties (Statistics Norway 2018). Moving from Oslo to the neighbouring county Akershus does not necessarily indicate increased labour mobility.
Assuming that immobile labour and nationally determined wages are the cause of regional unemployment, it would be profitable to subsidise labour, insofar as this compensates for the difference between the actual wage and the wage that would be derived from a free wage formation in the regional
The limited labour mobility in Norway may be exem-
labour market (L. Johansen 1965, NOU 1975: 2).
plifies by the response to declining oil prices, and following reduction in investment activity, in the petroleum industries in 2015. This caused nearly 8,300 employees to be laid off in the period from November 2015 to June 2016. Only 4 pct. of those who lost their job in the first half of 2016, and who were still living in Norway, moved to another county
In 2017 total unemployment was 4.2 pct., but with significant variation across municipalities.8 It seems that the unemployment rate increases with centrality (cf. Figure 2.3). However, the highest rates of unemployment occur more frequently among the most rural municipalities.
and few of them were from Rogaland where many became unemployed (Statistics Norway 2018). By November 2016 employment was higher among the few who moved from Rogaland, than those who
It cannot be ruled out that some share of the disability pensioners should be considered as a form of “hidden” unemployment. Looking at the share of disability pensioners we find little evidence that this is
6 7
18
This is further discussed in Chapter 5. Statistics Norway’s Labour Force Survey.
8
The total unemployment rate refers to unemployment in the Labour Force Survey, whereas municipal unemployment refers to registered unemployed. Thus, the average of the unemployment rates in Figure 2.3 is lower than 4.2 pct.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
a greater issue in more rural municipalities (cf. Fig-
2.2
Administration of the scheme9
ure 2.4). According to Chapter 23-2 of the National Insurance
Unemployment (%)
Figure 2.3 Unemployment1 and index2 of centralisation by municipality. November 2017
Act,10 all employers in Norway have a legal obligation to contribute to the national social security
10
scheme. The contribution is calculated as a share of
8
gross wages paid to the employees. The general rate in Norway is 14.1 pct. The regional (notified) aid
6
constitutes the reductions of the social security con-
4
tributions below the general rate. The tax rates are
2
determined annually by the Norwegian Parliament.
0
According to paragraph 12 of Chapter 23-2, the Par250
500
750
1000
Centrality Index Zone 1 Zone 4
Zone 1a Zone 4a
Zone 2 Zone 5
liament may adopt regionally differentiated contribution rates, as well as specific provisions for employ-
Zone 3
1) Registered unemployed 15-74 years 2) Ranging from 295 (lowest centrality) to 1,000 (highest centrality). Oslo is assigned the highest value. Source: Statistics Norway
ers within certain sectors.
2.2.1 Eligible recipients Prior to 2007 the tax rate for each employee was determined by the residence of the employee. After
Figure 2.4 Disability pensioners1 and index2 of centralisation by municipality. November 2017
rates vary according to where the firm is located. The employer (firm) is automatically entitled to the
25
Disability pensioners (%)
2007 differentiated payroll taxes implies that the
reduced rate, i.e. no application is required. If the
20
firm has establishments with different addresses, a
15
reduced tax rate only applies to employees who
10
work within the eligible area (see Chapter 2.4 for de-
5
scription of eligible areas). If employees spend half or more of their working time in a tax zone other than
0 250
500
750
1000
rate is based on the applicable rate in the zone in
Centrality Index Zone 1 Zone 4
Zone 1a Zone 4a
Zone 2 Zone 5
the one in which their employer is located, the tax
Zone 3
which the employees spend most of their time.11
1) As a percentage of the population 18-67 years 2) Ranging from 295 (lowest centrality) to 1,000 (highest centrality). Oslo is assigned the highest value. Source: Statistics Norway
9
This section is based on EFTA Surveillance Authority decision of 18 June 2014 on regionally differentiated social security contributions 2014- 2020 and mainly describes the scheme as notified for the period 2014-2020. Changes in rates and eligible areas are presented in the next section. 10 LOV-1997-02-28-19.
11
From 1 January 2016, it was no longer possible for employers with ambulatory activities to pay a lower rate than the rate applicable to the zone in which the firm had its address.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
19
2.2.2 Sectoral exceptions
annual budgets (cf. Figure 2.5). Since 2007, private
Firms operating in the following sectors or activities
sector has accounted for almost 60 pct. of the esti-
are not eligible for aid (reduced tax rate) under the
mated forgone tax revenues.
scheme:12 Forgone tax revenues are calculated as the differa.
Steel13
b. Synthetic c.
ence between the potential tax revenue if all firms fibres14
faced a payroll tax rate of 14.1 pct. and what is paid
Transport15
with differentiated rates. Thus, increased wages,
d. Airports16 e. f.
and increased employment, is the main explanation
Energy17 Financial and insurance
for the increase in annual budgets. It is worth noting activities18
g. Head office and consultancy
activities19
that the calculation of annual forgone tax revenue assumes everything will continue as is if the differentiation was removed, i.e. assuming no firm clo-
Firms with activities both inside and outside the
sures and no layoffs. 20
scheme will be eligible for a reduction in the payroll activities. However, this requires keeping separate accounts, clearly identifying direct and indirect labour costs and allocating them based on consistently applied and objectively justifiable principles, to demonstrate that the ineligible activities will not benefit from a reduced tax rate. Firms with outstanding recovery orders and firms in difficulties will not be eligible for aid under the
Figure 2.5 Estimated loss of revenue (tax relief). NOK billion. Constant 2017 prices. 2000-20181,2 18 16 14 12 10 8 6 4 2 0
scheme.
Private sector share (right axis)
2.2.3 Annual budget
Total tax relief (left axis)
The scheme had a budget of about NOK 13.5 billion in 2017 (estimated loss of tax revenues). Apart from years with restrictions in the scheme (due to ESA regulations), there has been a steady increase in
12
As of 1 January 2018, firms operating within the transport and energy sector are eligible for reduced tax rates. 13 As defined in Annex IV of Guidelines on regional State aid for 20142020 (p. 43). For the purpose of the evaluation we have defined the steel sector as NACE Rev. 2 group 24.1. 14 As defined in Annex IV of Guidelines on regional State aid for 20142020 (p. 43). For the purpose of the evaluation we have defined the synthetic fibres sector as NACE Rev. 2 groups 13.1, 13.2 and 13.3. 15 NACE Rev. 2 classes 49.100, 49.200, 49.311, 49.312, 49.391, 49.392, 49.393, 49.410, 50.101, 50.102, 50.109, 50.201, 50.202, 50.203, 50.204, 50.300, 50.400, 51.100, 51.210
20
90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
tax for the labour costs strictly related to the eligible
1) Proposal for 2018. 2) Budgets prior to 2004 do not distinguish between loss of revenue from private and public sector. Source: Ministry of Local Government and Modernisation (annual budget proposals)
16
See Guideline on regional State aid for 2014-2020 (p. 3). NACE Rev. 2 division 35 18 NACE Rev. 2 division 64, 65 and 66 (Section K) 19 Undertakings performing intra-group activities and whose principal activity fall under NACE Rev. 2 classes 70.10 or 70.22 20 This may be likely in the long run; those who lose their jobs due to closures or downsizing get another job. In the short run, however, unemployment might result in lower tax revenues than what would be the case if all existing firms paid a tax rate of 14.1 pct. for all existing employees. 17
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
Figure 2.6 Payroll tax rates by tax zone. 1975-2017 20%
20%
18%
18%
16%
16%
14%
14%
12%
12%
10%
10%
8%
8%
6%
6%
4%
4%
2%
2%
0% 0% 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 Zone 1
Zone 1a
Zone 2
Zone 3
Zone 4
Zone 4a
Zone 5
Note: Zone 1a was introduced in 2007 with the same tax rate as Zone 2 up to a threshold (see Figure 2.8). Source: Statistics Norway
Figure 2.7 Share of municipalities by tax zone. 1975-2017 100%
100%
90%
90%
80%
80%
70%
70%
60%
60%
50%
50%
40%
40%
30%
30%
20%
20%
10%
10%
0%
0%
Zone 1
Zone 1a
Zone 2
Zone 3
Zone 4
Zone 4a
Zone 5 Source: Statistics Norway
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
21
2.3
Changes in the scheme
changes, ESA approved the Norwegian scheme, partly due to a flexible interpretation of rules for
Effective from 1 January 1975, the payroll tax was
transportation support in ESA and the Commis-
differentiated in three zones with tax rates varying
sion’s regional aid guidelines. In 2000, the justifica-
from 14 pct. in Zone 3, 16 pct. in Zone 2 to 17 pct.
tion of the scheme was changed to supporting firms
in Zone 1. The tax rate per employee was initially
through reduced payroll taxes to compensate for
dependent on where the employees lived. Thus,
travel distance in sparsely populated areas. The
firms hiring from different tax zones, faced different
scheme was thus considered to be operating aid in
costs on potentially equal labour.
accordance with the EEA State aid rules and approved as an indirect transport aid scheme.
A fourth zone (which today is Zone 5) was added to the scheme in 1981, with a tax rate of 8.6 pct. The
Effective from 1 January 2000, further changes in
tax rate was gradually reduced in this zone, and in
the scheme led to 53 municipalities changing tax
1990 another tax zone was added with a tax rate
zone. In total 39 municipalities faced lower tax rates
between Zone 3 and (the now) Zone 5 (cf. Figure
(most of them moving from Zone 2 to Zone 3),
2.6). From 1991 the tax rate in Zone 5, also called
whereas 14 municipalities moved to a zone with
the “Action Zone”, has been zero. The Action Zone
higher rates (from Zone 2 to Zone 1).23
covers all municipalities in the county of Finnmark in addition to seven municipalities in Nord-Troms
In 2002 new rates were introduced for employees
(i.e. the northernmost part of Norway, marked in red
who were 62 years and older and who were obli-
in Figure 2.9).
gated to pay taxes. This was put in place to stimulate employment of workers who might otherwise re-
From 1990 to 2007 municipalities were divided into five different payroll tax
tire. It was, however, removed in 2007.
zones.21 The tax rate increased in Zone 2, 3 and 4 for a short
In 1993, an additional tax was introduced for all em-
period in 2004-2006 due to EEA regulations. How-
ployees with earnings exceeding 16 times the basic
ever, in 2006 EFTA adopted new Regional Aid
(16G).22
Guidelines, which gave greater flexibility to grant
When introduced, this tax rate was 10 pct. on the
state aid in the least populated regions (EFTA
amount above 16G, regardless of tax zone. It was
Surveillance Authority 2006). Hence, payroll taxes
increased to 12.5 pct. in 1998. This scheme lapsed
were again decreased in the three zones in 2007. In
in 2006.
addition, the scheme was extended to seven zones
amount in the National Insurance Scheme
(adding Zone 1a and 4a).24 In 1999, the EFTA Court ruled that regional differentiated social security contributions implied illegal
During the period of increased tax rates between
state aid. However, later that year, following several
2004 and 2006, firms in the affected tax zones only
21
23
We have omitted to elaborate on a temporary experiment with six tax zones in 1990 and 1991. This has no practical meaning for the review of the scheme or the empirical analysis. Norwegian readers are referred to Helde (1998) for an elaboration of this period. 22 Equalled 16 x NOK 37 300 (yearly amount) in 1993.
22
This reform is described in more detail in our empirical approach in Chapter 4.1. 24 This reform is described in more detail in our empirical approach in Chapter 4.2.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
faced the higher tax rate on labour cost above a
2.4
Seven different tax zones
threshold. This still applies for firms in Zone 1a. That is, when labour costs exceed the threshold, the firm
Since 2007 Norway has been divided in seven dif-
faces a higher tax rate on the amount above the
ferent tax zones (cf. Figure 2.9). The designation of
threshold. In 2018 the tax-deductible amount is
eligible areas is guided by the principles that (i) the
NOK 500 000, which corresponds to following la-
relevant region should face a real need for regional
bour cost threshold
aid and that (ii) regions facing similar challenges should be treated equally (EFTA Surveillance
500 000 𝑙𝑐𝑜𝑠𝑡 ≥ ≅ 14.3 𝑚𝑖𝑙𝑙. 0.141 − 0.106
Authority 2014).
There have been several changes in the labour cost threshold since 2004 (cf. Figure 2.8).
Though the eligible municipalities account for more than half of all municipalities, they only account for one fifth of the population. Further, the tax zones
Figure 2.8 Tax-deductible amount and corresponding wage cost threshold. 2004-2018
consist of municipalities which vary greatly in both size and development in central characteristics
600
12
400
8
200
4
0
0
Millioner
16
municipality’s degree of regional disadvantages, the authorities has developed a periphery index.
Figure 2.9 Municipalities by payroll tax zone. 2018
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Tusener
such as population and employment. To assess a 800
Tax-deductible (left axis) Labour cost threshold (rigth axis) Source: The Norwegian Tax Administration
From 1 January 2007, the determination of the employees’ payroll tax rate changed from their place of residence to the location of the firm. Further changes were made in July 2014, relocating 31 municipalities to zones with reduced tax rates. In addition, selected sectors and activities, regardless of municipality, were no longer eligible for aid under the scheme (EFTA Surveillance Authority 2014). These changes still apply. They are, however, not
Source: Statistics Norway Map: ©Kartverket
part of the current evaluation due to data limitations.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
23
The Ministry of Local Government and Modernisa-
the same geographical disadvantages as municipal-
tion revised the periphery index in 2018, which now
ities covered by the scheme (cf. Figure 2.10).
comprises three indicators intended to reflect geo-
centrality index captures the geographic disad-
1000
vantages, whereas growth in employment (eco-
1200 800 1000 600 800
indicators behind the periphery index.
0
2.4.1 Geographic disadvantages
Zone 5
teristics of the seven tax zones briefly, including the
Zone 4a
The next sections discuss some important charac-
600 400 400 200 200
Zone 4
Government and Modernisation 2018).
Zone 3
captures societal challenges (Ministry of Local
Zone 1
nomic growth) and population growth (demography)
Zone 2
sulting from these disadvantages; the municipality’s
Figure 2.10 Municipalities by payroll tax zone and centrality index. 2017
Zone 1a
graphical disadvantages and societal challenges re-
Source: Statistics Norway
Geographic disadvantages are associated with small local and regional labour and service markets
Given the high weight on centrality, the distribution
with long distances to larger and more specialised
of municipalities along the periphery index is almost
markets. These disadvantages are captured in the
identical to the distribution of municipalities along
so-called centrality index, which is calculated based
the centrality index (only scaled from 0 to 100).
on the number of workplaces and different service
Thus, municipalities in different tax zones may have
functions that can be reached by car within 90
the same periphery index. However, the periphery
minutes
index is mainly used to assess regional disad-
(Ministry of
Local Government and
Modernisation 2018, Høydahl 2017).
vantages and is not the only determinant of the municipalities’ tax zone. Another important parameter
The centrality index takes on values from 295 to
when determining the eligible geographic areas is,
1,000 along a continuous scale, where the most
according to ESA’s Regional Aid Guidelines (RAG),
central municipality (Oslo) is assigned the highest
population density (EFTA Surveillance Authority
value.25
2014).
In the calculation of the periphery index, the
centrality index is given the most weight (60 of 100). Municipalities in Zone 1 faces the general payroll
2.4.2 Population growth and density
tax rate of 14.1 pct. and is outside the geographical
The objective of the scheme is to reduce or prevent
scope of RDSSC. However, measured by the cen-
depopulation in the most sparsely populated re-
trality index, several of these municipalities faces
gions in Norway. Municipalities facing the general payroll tax rate of 14.1 pct. (Zone 1) covered 78 pct.
25
24
In theory the index can go from 0 to 1,000.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
of the Norwegian population in 2014 (the last year
The overall population density in Norway is a little
in our evaluation period), whereas slightly less than
over 17 people per square kilometre, which under-
2 pct. of the population lives in Zone 5, where the
lines that large parts of Norway is not populated.
payroll tax rate is zero (cf. Table
2.1).26 All tax zones but Zone 3 has experienced positive
Table 2.1 Population growth and density in the seven different tax zones
Zone Zone 1 Zone 1a Zone 2 Zone 3 Zone 4 Zone 4a Zone 5 Norway
Share of population 2014 78.0 % 2.8 % 6.7 % 2.1 % 6.2 % 2.4 % 1.8 % 100 %
Population Annual popudensity lation growth 2014 2004-2014 78.6 1.4 % 9.4 0.6 % 4.7 -0.1 % 2.5 -0.3 % 5.1 0.0 % 32.0 1.4 % 1.7 0.2 % 16.9 1.1 %
Note: Population density measured as people per square kilometre of land area. Annual population growth is calculated as average annual growth in total population within each zone. Source: Statistics Norway
average annual population growth the last decade (cf. Table 2.1). This tax zone covers the most peripheral areas of Southern Norway and is largely consisting of mountain areas (Ministry of Finance 2014).
2.4.3 Employment and wage growth Employment (measured by place of work)27 is distributed between the seven tax zones with similar shares as the population. The share of employees in Zone 1 is almost identical with the zones’ share of the population, i.e. almost eight out of ten jobs are in Zone 1. The variation in employment growth between the
Apart from Zone 4 and 4a, the tax rate corresponds to the tax zones’ population density, i.e. the tax rate is lower in zones with lower population density. Zone 4a has the second highest population density, though the tax rate is lower than in both Zone 1a and 2. Zone 4a consists of only two municipalities (Tromsø and Bodø), both of which are larger cities (in Norwegian context). However, they are both surrounded by more sparsely populated areas and it is argued that a higher tax rate in Tromsø and Bodø
different tax zones is somewhat smaller than the variation in population growth (cf. Table 2.2). Average annual employment growth was 1.5 pct. in the period 2004-2014.28 Employment growth is highest in the most central tax zones, including Zone 4a (covering the two cities Tromsø and Bodø). There is a clear relationship between employment and population growth, though there is no clear answer to whether people follow jobs or the other way around (see Chapter 5).
could have undesirable effects on the neighbouring regions (Ministry of Finance 2014). Municipalities in Zone 4 are characterised by their remoteness from central markets and travel distance to Oslo.
Hourly wages are highest in the most central tax zones. This may be explained by the fact that people with high levels of education, and hence relatively high wages, tend to be attracted to cities with
26
These shares are almost identical in 2018. To give an indication of the number of existing jobs in the various tax zones. 27
28
Starting in 2015, the employment statistics are based on new data. Thus, employment figures are not comparable before and after this change.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
25
urban qualities. These mechanisms are discussed in more detail
later.29
ment rates in the two former zones may be explained by immigration and national migration patterns.
Table 2.2 Employment and wage growth in the seven different tax zones Annual employ- Average hourly ment growth wage growth Zone 2004-2014 2004-2014 Zone 1 1.7 % 4.3 % Zone 1a 1.6 % Zone 2 0.3 % 4.3 % Zone 3 0.6 % 4.3 % Zone 4 0.7 % 4.4 % Zone 4a 1.4 % Zone 5 0.8 % 4.9 % Norway 1.5 %
Mean unemployment rate 2004-2014 2.0 % 1.6 % 1.7 % 1.4 % 2.0 % 1.9 % 2.8 % 2.0 %
Note: Population density measured as people per square kilometre of land area. Annual employment growth is calculated as average annual growth in total population within each zone. Growth in hourly wages is based on the sample used in Chapter 4.3 (Zone 1a and 4a did not exist in 2004). Unemployment is calculated as total number of unemployed (15-74 years) divided by total population (15-74 years) in each zone. Source: Statistics Norway
Wages in Norway is to a large degree determined through centralised negotiations, which is reflected
2.4.4 Establishments Jobs can be created by expanding existing firms or establishing new firms. Establishments is highly concentrated in Zone 1 (cf. Table 2.3). About 80 pct. of new firms are established in Zone 1, of which almost 40 pct. are established in Oslo and Akershus.
Table 2.3 Establishment of firms in the seven different tax zones
Zone Zone 1 Zone 1a Zone 2 Zone 3 Zone 4 Zone 4a Zone 5 Norway
roll tax) differs somewhat from the other zones. Employees in Zone 5 has had the highest growth in average (and median) hourly wages (for full-time employees) in the period 2004-2014 (cf. Table 2.2). There is no systematic pattern in unemployment, considering the tax rates the different zones are facing. Zone 5 has the highest share of unemployed and higher than the overall unemployment rate (cf. Table 2.2). Apart from Zone 1 and Zone 1a, unemployment has decreased in all tax zones after the global financial crisis. Some of the rise in unemploy-
Annual growth in establishments 2004-2014 2.4 % 1.9 % 2.5 % 1.1 % 2.0 % 2.3 % 1.5 % 2.3 %
Source: Statistics Norway
in the small differences in wage growth between the different tax zones. However, Zone 5 (with zero pay-
Establishments 2014 48,312 1,270 3,090 802 2,726 1,230 744 58,174
All tax zones, as well as most municipalities within the zones, has experienced growth in the number of new firms between 2004 and 2014. Zone 4a, facing a tax rate of 7.9 pct., has experienced growth on level with Zone 1 in which firms face the general tax rate of 14.1 pct. However, it is worth noting that most newly established firms do not have employees and therefor do not pay payroll taxes, regardless of tax zone. It is further worth mentioning that for most figures discussed above there is variation (sometimes significant) between municipalities within the same tax zone.
29
26
Chapter 5 discusses the dynamics of regional population growth.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
3
Theoretical framework
Under the assumption of perfect competition, in-
sion is changed, and could thus be changed to-
cluding no economies of scale, prices of tradable
wards the solution one would get without centrally
products and mobile factors of production will be
determined wages.
equal
everywhere.30
Conversely, the prices of prod-
ucts and factors of production that are not mobile
A widely used argument against labour subsidisa-
may wary between different geographical locations.
tion is that it ultimately leads to lower capital inten-
The economy is characterised by optimal allocation
sity, which in turn leads to lower productivity and, in
of resources, i.e. there will be nothing to gain from
the longer term, welfare losses. However, in the
reallocating resources within existing production
case with centralised wage bargaining the effect of
processes, to production of other goods or services
reducing local labour cost is to counteract devia-
or to other regions.
tions between market earnings and actual wages that exists in the first place. By reducing labour cost
In practice, the mobility of labour is limited, while
through reduced payroll tax, the difference between
capital mobility is high, especially in the long run. In
the national wage and the locally optimal wage can
Norway, wages are to a large extent determined in
be removed, which in turn leads to a more optimal
centralised wage negotiations. This leads to a rela-
resource allocation, e.g. higher employment. How-
tively high degree of wage equalisation for equal
ever, the strength of this (direct) effect on employ-
work between geographical regions. Thus, wages
ment depends on to what degree the reduction in
(and prices of capital) will not perfectly reflect the
labour cost is transferred to higher wages.
scarcity of production factors. This may lead to higher wages and lower employment than what is
The regionally differentiated payroll tax in Norway
implied by “the free market solution”. This could typ-
was introduced in line with economic theory, show-
ically be the case in more remote areas with small
ing that labour subsidies would be better suited than
labour markets and/or a one-sided industrial base.
capital subsidies to stimulate regional employment in an economy with high capital mobility, low labour
Demand for labour will in general depend on the
mobility and a national collective wage bargaining
profitability of the firm, not the overall social welfare.
system.
From the firm’s point of view, it is profitable to employ labour up to the point where the value added of
In the following, we will discuss theoretically how
the last hour worked equals the hourly wage. The
employment and wages may react to a change in
firm’s volume of production and composition of la-
the pay roll tax under alternative assumptions. We
bour and capital in the production depend on input
use a stylised framework suited to illustrate the main
prices (wages and interest rates) and on the mar-
mechanisms at work.
ginal income (which depends on the properties of the demand curve facing the firm). When the rela-
From the theoretical discussions below, we derive
tive prices of factors of production is changed, e.g.
several hypotheses we wish to test in our empirical
due to lower payroll taxes, the firm’s optimal deci-
analysis in Chapter 4, i.e. effects of changes in the
30
See footnote 2.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
27
payroll tax on wages, employment, capital services,
of one factor changes without any impact on the de-
value added and establishment of firms (and exits).
mand for the other, the factors are said to be independent.
3.1
Demand for labour Following these examples, the net effect on the de-
Standard textbook micro economics, assuming all
mand of capital, from a reduction in the labour cost,
firms are profit maximising and can employ as many
is positive if the two factors are complementary fac-
workers as they wish at the going market wage rate,
tors of production. That is, the production requires
implies that a relative reduction in the price of a fac-
an increase in capital along with the increase in the
tor of production will increase a firm’s preferred use
use of labour.31
of this factor. This is easily demonstrated in the case of a firm utilising two inputs (without loss of general-
In addition to the case of complementarity, market
ity), labour and capital. If the payroll tax is reduced,
imperfections could lead to increasing capital in-
so is the cost of labour (relative to capital), and firms
vestments from reduced payroll taxes. If a firm
will switch towards a more labour-intensive produc-
wishes to increase production when the labour cost
tion. The effect on capital demand is not so clear; it
is reduced, but additional labour is not available, in-
is determined by the net effect of a positive income
vesting in labour saving technologies could be an
effect (with lower labour costs the firm can increase
option. Another possibility is that credit restrictions
production and, thus, the use of capital without in-
have been limiting the firm’s investment possibili-
creasing total costs) and a negative substitution ef-
ties, and that a lower payroll tax releases funding for
fect (capital has become relatively more expensive
capital investments.
than labour). The net effect on capital is therefore an empirical question.
The different effects of a change in price of labour is illustrated in Figure 3.1. The initial (optimal) compo-
Three simplified, yet enlightening, examples from economic theory are: (i) if an increase in the price of
sition of labour and capital of a firm is given by point 𝑋, where the budget constraint (line 𝐵1 ), indicating
use of the other, ceteris paribus, we say that the fac-
feasible combinations of labour and capital at a given cost, and the isoquant (curve 𝐼1 ), indicating all
tors have positive cross-price elasticity and are al-
factor combinations that produce the same amount
ternative factors of production or substitutes; (ii) if
of product, are tangent. In this case the optimal level og labour and capital is 𝐿1 and 𝐶1 , respectively.
one factor of production leads to an increase in the
an increase in the price of one factor leads to a reduction in the demand for the other, the two factors have a negative cross-price elasticity and are com-
A reduction in the pay roll tax implies that the firm
plementary factors of production, i.e. the factors are
can employ more labour without increasing costs, and the budget constraint shifts to 𝐵2 . If the firm,
mutually dependent in the production; (iii) if the price
however, keeps production at the same level as before the reduction in the payroll tax, i.e. reduce their
31
The net effect can also be positive in a case of alternative factors of production if the positive income effect outweighs the negative substitution effect.
28
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
costs for a given production, the new budget constraint is illustrated by the dashed line 𝐵𝑌 , with the
The effect on labour demand of lower labour costs
same slope as 𝐵2
Then the optimal use of labour
downward sloping demand curve in a wage-labour
and capital would be where 𝐵𝑌 is tangential to 𝐼1
diagram (see Chapter 3.3). In an “opposite” case of
(point 𝑌), i.e. increasing the use of labour to 𝐿𝑌 and
reduced cost of capital (e.g. due to capital subsi-
reducing the use of capital to 𝐶𝑌 . This is the substi-
dies), there would be an unambiguously positive ef-
tution effect; capital is relatively more expensive and
fect on capital and an undecided net effect on la-
the demand for capital is reduced.
bour.
Figure 3.1 Demand for labour and capital. Income and substitution effects of reduced labour costs
The magnitude of the effects depends on the slope
.32
is always positive. This may also be illustrated by a
of the budget constraints, i.e. the relative price of labour and capital, the size of the price change and the shape of the isoquant. The latter is determined by the degree of substitutability between the two factors of production.
3.2
Supply of labour
In the previous section, we discussed the demand for labour and capital (in partial equilibrium). To illustrate the total effect on a regional labour market of a change in relative factor prices, we need to introduce the supply side. Standard microeconomic theory for the labour market assumes that people are rational and maximise their utility in a trade-off between positive preferences for leisure and income resulting from time spent working. If the inThe income effect follows from the budget con-
come is spent in full on consumption, the trade-off
straint moving outwards, indicating the new feasible level of production (𝐼2 ) with the same costs as be-
is between leisure and consumption.
fore the change in the payroll tax. The new budget constraint (𝐵2 ) if tangent with 𝐼2 at point 𝑍, and the
This is illustrated in Figure 3.2, where the line 𝑀1
new composition of labour and capital is indicated by 𝐿2 and 𝐶2 , respectively. Note that the net effect
is spent on leisure there is no consumption and,
on capital is positive in this case. This follows if the
are spent working. Every additional hour of leisure
income effect is greater than the substitution effect,
must be met by an equal reduction in hours worked
represents the initial budget constraint; if every hour conversely, consumption is maximised if all hours
which need not be the case.
32
The slope of the budget constraint equals the ratio between the prices of the two input factors, i.e. the market-exchange ratio between labour and capital.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
29
and a corresponding loss of income (and consump-
same level as before the change in the wage. With
tion). All combinations of leisure and consumption
a higher hourly wage rate, each hour with leisure is
generating the same level of utility is represented by an indifference curve, where 𝐼𝐶1 is the indifference
now relatively more expensive than before (the al-
curve defining maximum achievable utility given the budget constraint 𝑀1 .
with a higher wage). Thus, the substitution effect is negative (𝑋𝐶 < 𝑋𝐴 ). However, with a higher wage it
ternative to one hour of leisure is one hour of work
is possible to increase the utility without increasing The initial utility maximising combination of leisure and consumption is defined by point 𝐴, giving lei-
the number of hours worked. This is the income effect (parallel shift of the dashed line to point 𝐵).
sure 𝑋𝐴 and consumption 𝑌𝐴 . If a reduction in the payroll tax is partly passed over to increased wages,
With an increase in the wage rate, the net effect on
possible consumption with no leisure increases from 𝑌1 to 𝑌2 , as income increase. The new budget
consumption is unambiguously positive; income is
constraint (𝑀2 ) is tangent with an indifference curve
the case drawn in Figure 3.2 the net effect on leisure
with higher utility in point 𝐵. Again, the move from 𝐴 to 𝐵 can be decomposed in two separate effects, an
(work) is negative (positive), compared to the initial adjustment (point 𝐴). As discussed in the previous
income effect and a substitution effect.
section, this may not be the case. Thus, the net ef-
higher for all possible choices of hours worked. In
fect on labour supply of a wage increase is uncer-
Figure 3.2 Supply of labour. Income and substitution effects of increased wage
tain. Individual supply curves may, under standard assumptions, be aggregated to a macro supply function. Given that the substitution effect is greater than the income effect (as above), the aggregated labour supply curve is upward-sloping in a wage-labour diagram (cf. Figure 3.3).
3.3
Market equilibrium
In this section we combine the labour demand and supply from the two previous sections in a labour market model to illustrate the total effect on employment and wages of a reduction in the pay roll tax, as well as how the effect may depend on the demand and supply elasticities (i.e. the slopes of the two curves). The substitution effect is illustrated by the dashed line, tangent with the initial indifference curve (𝐼𝐶1 )
With a strictly negative relationship between wage
in point 𝐶. The corresponding levels of leisure and
costs and demand for labour (labour demand de-
consumption, 𝑋𝐶 and 𝑌𝐶 , is the optimal allocation
creases with increasing wages), the demand curve (𝐷) is downward-sloping in a wage labour diagram
with a higher wage, but keeping the utility at the
30
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
(cf. Figure 3.3).33 The less sensitive – or inelastic – demand is to a change in wages, the steeper the
In both markets, the market clearing wage is indicated by 𝑤 ∗ , with corresponding employment 𝑙 ∗ . To
slope of the demand curve.
illustrate possible effects of a reduction in the payroll
Assuming labour supply increases with an increase
tax on both wages and employment, we start with an initial payroll tax equal to 𝑡1 . At this tax rate the
in real wages (as shown in the previous section), the
demand curve is shifted downwards (not shown in
labour supply curve is upward-sloping. The slope of the curve is determined by the labour supply elas-
the graphs) and intersect the supply curves in point 𝑏. The wage payed to the employees is given along
ticity.
the y-axis by the dashed line from point 𝑏 and supply of labour, at this wage, along the x-axis in the
In equilibrium, demand equals supply. For a given
same point. Both the wage payed to the employees
demand curve, the equilibrium wage depends on
and the supply of labour is lower than without a pay-
the elasticity of labour supply, as illustrated in Figure 3.3. In market 𝐴, supply is relatively elastic, i.e. a
roll tax. The employers’ wage cost is, however, higher than 𝑤 ∗ , given along the y-axis by the dashed
wage increase causes a relatively large increase in the supply of labour. In contrast, market 𝐵 illustrate
line from point 𝑎.
the case of inelastic supply, where a wage increase lead to a relatively small increase in the supply of
The tax wedge after a reduction of the pay roll tax is illustrated by 𝑡2 . The resulting increase in employ-
labour.
ment in the two segments are shown by 𝐿1 and 𝐿2 , respectively. With an equal change in the payroll tax
Figure 3.3 Labour market equilibrium with different supply elasticities
33
Strictly negative implies that the effect on labour demand of increased wages cannot be zero.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
31
(from 𝑡1 to 𝑡2 ) and a given demand, the employment
pay roll tax would be a less effective measure to in-
effect depends on the elasticity of supply and is larger in market 𝐴 where supply is more elastic, i.e.
crease employment.34
where supply is more price sensitive.
3.4
With elastic labour supply, most of the tax incidence
The Norwegian labour market differs significantly
resides with the employers, i.e. the difference be-
from the case of perfect competition, as depicted
tween the market clearing wage and the wage cost,
above, which does not fully consider modifications
including payroll tax, is larger than the difference be-
caused by centralised wage negotiations. This is
tween the market clearing wage and the wage
non-negligible as close to half of Norwegian workers
payed to the employees. With inelastic supply it is
are organised (see NOU 1996: 9 for a discussion).
the opposite. Thus, for a given change in the payroll
It is likely that a reduction in the pay roll tax is more
tax, the reduction in wage cost per unit of labour (given by the distance 𝑤𝑐) is larger and the increase
efficient when wage negotiations are centralised
in wage payed to employees per unit labour (𝑤𝑤)
Mellander and Öckert 2009).
Effects with centralised wage negotiations
(Cappelen
and
Stambøl
2003,
Bennmarker,
smaller in market 𝐴 than in market 𝐵. A region-specific reduction in the pay roll tax will to The elasticity of demand (the slope of the demand
a lesser extent lead to a region-specific wage in-
curve) could, off course, also vary. One can imagine
crease when wage growth is regulated by central-
that demand for workers with high education and
ised agreements. Lower wage costs increase com-
skills could be relatively inelastic, assuming it is
petitiveness and makes it possible to increase pro-
harder to substitute high-skilled labour with low-
duction and the use of relatively less expensive fac-
skilled workers or machines (capital). This illustrates
tors of production, in this case employment. This
the importance of the industrial base for the effect of
may explain why empirical studies tend to find
a change in the payroll tax. A steeper demand curve
higher employment effects in the Nordic countries,
(inelastic demand) would lead to a lower employ-
and in Norway in particular. Alternatively, the firm
ment effect, smaller increase in the after-tax wage
could use the gain from reduced labour costs in lo-
payed to the employees and larger reduction in the
cal wage negotiations to attract more high-skilled
employer's labour costs per unit of labour.
workers.
To sum up, the model above illustrates effects un-
With nationally determined wages, there may be re-
der the idealised conditions of perfect competition
gional discrepancies between supply and demand
and predicts a positive effect on employment and
for labour. Figure 3.4 illustrate two such cases; one
wages of a reduction in the payroll tax. The magni-
with a regional supply surplus (i.e. regional unem-
tude of the effects depends on the elasticity of sup-
ployment) and one with a regional demand surplus
ply (and demand), which may vary between differ-
(labour shortage). The nationally determined wage 𝑊 𝑁 is higher than the market clearing wage in panel
ent segments of the labour market and regions. In regions where labour supply is inelastic, reduced
34
With inelastic supply more of the tax reduction is shifted on to higher wages, which in turn may have an indirect effect on labour demand
32
through increased household demand for goods and services. This is discussed in Chapter 3.5.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
𝐴 (representing rural regions). Conversely, in cen-
firms, respectively. Prior to the reduction in the pay-
tral regions (panel 𝐵) the market clearing wage is higher than 𝑊 . Differences in productivity, e.g. due
roll tax, unemployment is given by the gap between regional supply (𝐿𝑆1 ) and demand (𝐿𝐷1 ) of labour,
to long travel distances in sparsely populated re-
equal to 𝑢1 . Removing the payroll tax reduces the
gion, may explain the differences in the market
labour cost per unit of labour from 𝑊𝐶 𝐴 to 𝑊 𝑁 (i.e.
clearing wage in the two regions, all else equal.
labour cost equals the wage payed to the employ-
With equal payroll tax, 𝑡, in both regions the wage
ees).35 With the decrease in labour cost, labour demand increases to 𝐿𝐷𝐴 and unemployment de-
cost per unit of labour is 𝑊𝐶. A national set wage
creases to 𝑢2 .
𝑁
above the equilibrium wage leads to regional unemployment in rural regions, irrespective of the size of the payroll tax (here illustrated by 𝐿𝑆1 > 𝐿𝐷1 ). In the
In the central region the payroll tax remains un-
opposite case it is created an excess demand for labour (𝐿𝑆2 < 𝐿𝐷2 ), as in panel 𝐵. With a high
Thus, reducing the payroll tax only in the market
enough payroll tax, the excess demand will diminish (a tax such that 𝐿𝑆2 = 𝐿𝐷2 ).
market clearing wage, increases total welfare com-
changed and in the short run, nothing changes. where the national determined wage exceeds the pared to the situation with equal tax rates in both regions.36
Panel 𝐴 illustrates the initial situation in the rural region, with nationally fixed wages and the short run
The above is somewhat hypothetical. Wages are, in
effects of a reduction in the payroll tax. The supply
most cases, a result of both central and local wage
and demand curves represent existing workers and
Figure 3.4 National wages and heterogenous labour markets
35
Assuming the only difference between labour cost and wage per unit of labour is the payroll tax.
36
Reducing the payroll tax with an initial demand surplus will only increase the surplus.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
33
negotiations, allowing for regional deviations and re-
Lack of direct effects on employment does not mean
duction in regional unemployment without reduction
that total (regional) employment cannot increase.
in the payroll tax. Furthermore, the above discus-
Higher disposable income (through higher wages)
sion is concentrated on short run effects. In the
for those already employed is likely to increase their
longer run demand for labour could increase
demand for (locally produced) goods and ser-
through new and in-migrating firms. On the supply
vices.37 Thus, higher wages may indirectly affect
side, growing in-commuting and in-migration must
employment. A reduction in factor prices may also
be considered in the longer term and inactive peo-
lead to increased operating profits, trigger new cap-
ple may become active. Long run effects are dis-
ital investments and dividends (cf. Figure 3.5). If we
cussed in more detail in the following section.
assume that both capital and workers are mobile, though not instantaneously, relocation of firms (and
3.5
Time and direct versus indirect effects
workers) to regions with lower payroll tax may also give a long-term (positive) effect on employment.
In the short run employees’ bargaining power may be weak and the main effect of a reduction in the
There may also be additional positive effects on em-
payroll tax is likely reduced labour costs. If so, the
ployment. So far, we have assumed price taking be-
effect on employment is likely to be relatively strong
haviour. Realistically, most industries are character-
in the short run as well. However, both theory and
ised by a degree of monopolistic competition. In that
empirical results on a national level show that, over
case firms will respond to a reduction in factor prices
time, bargained wage increases will counteract the
by a certain reduction in product prices, leading to
initial effect of a reduced tax rate and one may ex-
increased demand for their products. According to
perience little, or even no, direct effect on employ-
economic theory, monopolistically competitive firms
ment.
Figure 3.5 Direct and indirect effects of a change in payroll taxes
Source: Samfunnsøkonomisk analyse AS 37
Assuming the number of hours worked stays constant despite higher wages.
34
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
normally respond to increases in demand by in-
The second equation, (3.2), is a product demand
creasing their demand for employment and other inputs.38
function with conventional assumptions about the partial derivatives, where 𝑃 41 and ̅ P is the product
If reduced payroll tax leads to higher employment
price and price on competing products, respectively, and 𝑌 is income.
and lower capital intensity than the optimal market solution, a too low capital intensity is associated
Equation (3.3) is a simple functional relationship for
with lower productivity and thus a welfare loss.
aggregate income in the geographic region we study, where CPI is the consumer price and 𝑇 is
3.6
transfers (alternative policy measures). For simplic-
A stylised model
To formalise the line of thought above, it may be
ity, we only consider wage income and transfers. If we apply the framework to a single firm (𝑁 is firm
useful to consider a stylised model for medium-run
employment), total employment is almost unaf-
employment determination, assuming that employ-
fected. That is, if only one firm face a reduced pay-
ment is determined from demand, i. e. that demand for labour is always accommodated by correspond-
roll tax the effect on employment through income is negligible, hence we set 𝛾 ≅ 0. Conversely, if the
ing supply adjusted for wage effects.39 For that pur-
change in the tax apply to all firms (𝑁 is regional
pose, we define the following textbook system of
employment), 𝛾 ≅ 1.
equations: Based on this framework the different effects of a 𝑊(1 + 𝜏) 𝑁 = 𝑁( , 𝐷) , 𝑄
𝑁1 ≤ 0,
𝑃 𝐷 = 𝐷 ( , 𝑌), 𝑃̅
𝐷1 < 0,
𝑊 𝑌 = 𝑌 (( ) 𝛾𝑁 + 𝑇), 𝐶𝑃𝐼
𝑌1 > 0
𝑁2 ≥ 0
(3.1)
reduction in the payroll tax can be expressed compactly as:
𝐷2 > 0
(3.2) 𝜕𝑁 −𝜕𝜏 𝜕𝑊 1 𝜕𝑃 1 𝛾𝑁 𝜕𝑊 (1 + 𝜏) + 𝑊) − 𝑁2 𝐷1 −𝑁1 ( − 𝑁2 𝐷2 𝑌1 𝜕𝜏 𝑄 𝜕𝜏 𝑃̅ 𝐶𝑃𝐼 𝜕𝜏 = 𝑊 1 − 𝑁2 𝐷2 𝑌1 𝛾 𝐶𝑃𝐼
(3.3)
Equation (3.1) gives the conditional demand for labour in the case of monopolistic competition, where 𝑁 is employment, 𝑊 is wage earnings per unit of labour, 𝜏 is social contribution taxation rate, 𝑄 is the
The first term in the numerator represents the effect of relative factor prices on conditional labour demand (direct effect). The effect is negative if 𝑁1 < 0
uct demand. Given the assumption of monopolistic
and is largest in absolute value when wage earnings ∂𝑊 are unaffected, = 0. As mentioned above, this is
competition, demand is set equal to output.
most realistic in the short run.
38
40
price of variable inputs in production40 and 𝐷 is prod-
If demand is not perfectly inelastic. As shown in Chapter 3.2, the slope of the supply curve is ambiguous, as it depends on the relative strength of the opposing forces of the substitution and income effect. It is common to assume that the substitution effect dominates, leading to an upward sloping supply curve (as discussed). However, for the sake of clarity, we simplify the supply side further in this section. 39
𝜕𝜏
In this setup we assume that there exist only two inputs of production; labour and another. The number of these may depend on the time horizon of the analysis, e.g. capital being fixed in the short-term analysis but variable in the long-term perspective. 41 With monopolistic competition P is a function of unit labour costs. For simplicity this equation is left out of the system.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
35
If the tax change is transferred to higher wages, 𝜕𝑊 then > 0. Theoretically, this may be the case if
pay roll tax on wages and employment (and in some
there is collective bargaining and firms and unions
industries, the organisation of the labour market and
have targets for their respective shares of value
wage formation in the different regions will affect the
added in the firms.
efficiency of the tax measure. Thus, based on
−𝜕𝜏
cases a ranking of the effects). The composition of
standard theoretical considerations, we cannot say The second and third terms represent the indirect
much about the magnitude of the effects. That re-
effect on employment through the effect on de-
mains an empirical question, which we will address
mand. The second term follows from the assump-
in the next chapter.
tion of monopolistic competition: monopolistic firms 𝜕𝑃 adjust their product price ( ≥ 0) to changes in la𝜕𝜏
bour costs and consumers change their demand to changes in consumer prices. The third term in the numerator illustrates that demand is increased if a change in payroll taxes is 𝜕𝑊 transferred to wage earnings ( < 0). Hence, the 𝜕𝜏
more wage earnings are affected, the more the effect through changes in relative factor prices (the first term) is moderated and the effect through changes in demand (the third term) is amplified. The denominator is always positive. It is less than one if 𝛾 is reasonably large, i.e. a reduction in firms’ employment has a numerically significant effect on the region’s total employment. With
𝜕𝑃 𝜕𝜏
> 0 (mark-up price setting due to monopo-
listic competition),
𝜕𝑊 𝜕𝜏
< 0 (changes in the payroll
tax affects wage earnings) and 𝑁1 < 0 (the direct price derivative of labour demand is negative) all three terms contribute to higher employment. Even if there are negligible possibilities of substitution, 𝑁1 ≈ 0 (may be realistic in the short run), there can still be effects on employment stemming from the two channels of increased product demand. The theoretical framework presented in this chapter helps understanding the central mechanisms and indicate the sign of the effects of a reduction in the
36
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
4
Empirical evidence
Using different well-established econometric ap-
common assumption in applied incidence studies of
proaches, we find evidence that changes in the pay-
payroll taxes is that the incidence is borne by the
roll tax to some degree are shifted onto workers, i.e.
workers (through decreased wages), regardless of
changes in the payroll tax affects wages. This holds
who has the legal obligation to pay the tax (i.e. the
for both reductions and increases in the tax rate.
statutory incidence). If this is the case, there is little
However, it seems that less of the tax is levied on
reason to believe that we will find effects on employ-
the employees in the case of an increase in the pay-
ment, precisely because the cost is shifted to the
roll tax than in the case of a decrease, indicating an
workers, who do not change their behaviour signifi-
asymmetry in adjustments to new tax rates. This is
cantly due to their relatively inelastic labour supply.
in line with what we would expect, especially in an economy with a relatively high share of organised
However, more recent studies find somewhat con-
workers and centralised wage negotiations. Further,
tradicting results. Saez, Matsaganis and Tsakloglou
the size of the tax incidence residing with the em-
(2012) use a reform of payroll taxes in Greece to
ployees seems to be sensitive to sample and model
study long-run tax incidence. They find that the em-
specifications, but in general we find that most of the
ployer-paid payroll tax fully resides with the em-
tax incidence resides with the employer. We further
ployer, whereas the employee-paid payroll tax re-
assess how the payroll tax (change in relative factor
sides with the employee. Thus, their results suggest
prices) affects firms’ labour demand, value added,
that employers do not pass on the extra cost of in-
capital investments and operating profit. We find a
creased employer payroll taxes to the employees.
positive effect on firm’s demand for labour, but the effects are moderate. Our findings are mostly in line
In 2007 (and 2009) Sweden introduced a country-
with previous studies of the regionally differentiated
wide lower payroll tax rate for young workers to fight
payroll tax.
youth employment.42 A recent study show that the payroll tax rate cut from 31 pct. down to 15 pct. for
The person who has the legal obligation to make a
worker 26 and younger had no effect on net-of-tax
tax payment may not be the person whose welfare
wages for the young workers, compared to slightly
is reduced by the presence of the tax. That is, the
older (untreated) workers, i.e. no tax shifting on to
economic incidence may differ from the statutory in-
employees. However, it seems that the reduced tax
cidence due to changes in behaviour and conse-
rate had positive effects on the employment rate of
quent changes in equilibrium prices (Fullerton and
the treated young workers, and larger in places with
Metcalf 2002). The theoretical discussion in Chapter
initially higher youth employment (Saez, Schoefer
3 illustrates how the economic incidence varies with
and Seim 2017).
supply and demand elasticities and how effects of changes in the payroll tax may differ between differ-
The statutory incidence of the Norwegian payroll tax
ent labour markets (regions).
is on the employer. Previous studies (e.g. Stokke (2016), Gavrilova, et al. (2017), Johansen and
There is a general expectation that labour demand
Klette (1997) find some shifting of the tax incidence
is more elastic than labour supply. Thus, the most
on to the workers through decreased wages, but not
42
The scheme was abolished in 2015.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
37
fully. That is, it seems that the employers do take
change in zones with lower rates in the absence of
some of the tax burden, at least in the short run. A
a differentiated payroll tax rate.
newly published discussion paper, exploiting the changes in RDSSC in 2004-2006, find that Norwe-
In this chapter we seek to overcome these chal-
gian firms are only partially able to shift the in-
lenges with different econometric approaches to as-
creased costs from higher payroll tax rates onto
sess the effects of changes in the payroll tax on
workers’ wages, and that firms in large respond to
wages, employment, value added and capital ser-
the tax increase by reducing employment (Ku,
vices in existing firms, as well entry and exits of
Schönberg and Schreiner 2018).
firms. We exploit the three reforms of the scheme that took place in the period 2000-2007 (see Chap-
The objective of the regionally differentiated payroll
ter 2.3): (i) we use difference-in-differences to study
tax is to reduce or prevent depopulation in the most
effects of a lower tax rate for firms in municipalities
sparsely populated regions in Norway by stimulating
that changed tax zone in 2000; (ii) we use both dif-
employment. Based on the abovementioned stud-
ference-in-differences and a regression kink design
ies, we expect to find effects on employment from
to evaluate effects of increased payroll taxes in the
changes in the payroll tax rate.
period 2004-2006; (iii) we exploit all variation in the tax rates following the reform in 2004 and its rever-
To identify the causal effects of the scheme, we
sion in 2007 to estimate long-run effects on the de-
must perform a counterfactual analysis, i.e. com-
mand for labour using a GMM estimator.
pare the actual level of any outcome variable with the level that would have been realised in the ab-
The tax rate in Zone 1 and Zone 5 has remained
sence of the differentiation. Ideally, the counterfac-
unchanged at 14.1 and 0 pct. throughout our evalu-
tual outcome should be determined by a controlled
ation period, respectively. Zone 1 is outside the ge-
experiment, randomly dividing the population of
ographical scope of the scheme and the lack of var-
firms into different groups facing different tax rates.
iation in the tax rate within this zone is thus not a
The effect of a higher (lower) tax rate could then be
problem. It is, however, challenging that we cannot
measured by the difference in response between
identify effects in Zone 5, which is the tax zone with
the groups with increased (decreased) labour costs
the largest difference between the actual and gen-
and the group facing the general tax rate (the pla-
eral payroll tax rate. To meet the latter challenge,
cebo group).
we present a descriptive analysis of how removing the differentiation in Zone 5 will affect the firms’ op-
A random experiment is obviously not possible.
erating profits.
Comparing firms from different tax zones with each other is also far from the golden standard of ran-
Table 4.1 summarise samples, econometric ap-
domly selected groups. The introduction of different
proaches and findings. A detailed presentation is
payroll tax rates and their adjustment over time
given in the following chapters.
were primarily intended to stimulate employment in rural areas and areas experiencing depopulation. Thus, we cannot use employment growth in tax zones with higher tax rates (experiencing less or no depopulation) to infer how much employment would
38
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
Table 4.1 Overview of the empirical results Effects Reform 2000: 53 municipalities changed tax zone, of which 34 municipalities from Zone 2 to 3 with reduced tax rate
Sample1 All workers and firms in selected municipalities in Zone 2 (control group) and Zone 3 (treatment group)
Wages Significantly higher wage growth in treatment group. Tax shifting onto employees around 24 pct. in the main specification (29 pct. in tertiary sector). The tax incidence residing with the employees varies between 0 and 53 pct. in other specifications.
All employees in firms in Zone 2, 3 and 4
2004-2006 RKD
2003-2014 FE, BE, GMM
Years Approach 1997-2003 Diff-in-diffs
2004: Increased tax 2000-2006 Diff-in-diffs rates in Zone 2 to 4 for firms with labour costs above a threshold
2004 and 2007: Variation in rates following changes in the scheme
Employment Positive short-term effect of 3-4 per centage points on employment growth of the reduced tax rate (both on the extensive and the intensive margin). The effect is driven by the tertiary sector.
Capital Positive effects in the secondary sector, dominated by manufacturing.
Value added Some positive effects, similar to the employment effects.
Significantly lower wage Not assessed. growth in Zone 2. In Zone 3 and 4 the effect only applies to firms reaching the highest threshold in 2004.
Not assessed.
Not assessed.
Selected firms in Zone 2 and 4
4-17 pct. of the increase in total wage costs is shifted onto workers in Zone 2 and 0.5-4 pct. in Zone 4.
Not applicable (too data demanding).
Not applicable (too data demanding).
Not applicable (too data demanding).
All firms in Zone 1a-4
0-29 pct. of increase in total wage costs is shifted onto workers (0 pct. with FE and GMM, 29 pct. With BE and 8 pct. with OLS).
Long-run labour demand elasticity (on the extensive margin) equal to -1.1
Not assessed.
Not assessed.
1) Detailed descriptions of the samples in the following chapters.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
39
4.1
Evaluation of the reform in 2000
In our evaluation of the effects of the regionally differentiated payroll tax rate, we rely on so-called exogenous shocks, or quasi- experiments, to employ
Table 4.3 Indicators and weights used as basis for the changes 1 January 2000 Indicator category
Indicator
Weight
Geography
Centrality
0.30
Population density Share of population residing in urban area
0.10
valid research methods and acquire results that satisfy certain methodical standards. One such shock is the change in the scheme that occurred in 2000 (see Chapter 2.3). Effective from 1 January 2000, 53 municipalities were included in different tax zones than prior to the change, facing new tax payroll tax rates. Of these, 32 municipalities were moved from Zone 2 to Zone 3, which meant a reduction in the payroll tax rate, while 14 municipalities were moved from Zone 2 to Zone 1, resulting in
0.15
Labour market Share of unemployed
0.10
Income
Zone 1 to Zone 2.
New zone Zone 3 Zone 1 Zone 4 Zone 2
0.05 0.05
0.25
0.05
0.15
Income per tax payer
0.10
0.10
Source: Mønnesland et al. (2002)
The indicators used to determine which municipalities should change zones, and the weights used,
Table 4.2 Changes in tax zones 1 January 2000 Initial zone Zone 2 Zone 2 Zone 3 Zone 1
0.5
Share on disability
an increase in the payroll tax rate. Further, six municipalities moved from Zone 3 to Zone 4 and one
0.10
Demographics Population change Population shares; everyone aged 18-30 Population shares; women aged 20-39
Category weight
No. of. municipalities 32 14 6 1
Change in tax rate (percentage points) -4.2 3.5 -1.3 -3.5 Source: Strøm (2002)
are listed in Table 4.3. Only using these indicators as basis for the differentiation of the payroll tax rate would lead to a messy and complex map of tax zones, and an inexpedient differentiation by municipalities that form larger economic and labour market regions. Therefore, a certain degree of discretion was used to adjust the indicator-based proposals for changes. The conservative party disagreed with the government’s proposal for changes,
The reform came as a response to a change in clas-
stating that the proposal was to a great extent based
sification of the scheme as a compensation scheme
on discretion, as opposed to more objective criteria.
for disadvantaged regions regarding transportation distances. The reclassification meant there was a
Even though the tax zone changes were endoge-
need to make some adjustments in some municipal-
nous in terms of the listed indicators, we argue that
ities’ tax rates. Purely geographical elements, like
the discretion used in the decisions of which munic-
centrality were given a large weight in the revisions.
ipalities that would change zones, and the large
Other indicators used as basis for the changes, but
weight given to geographical indicators, means the
given less weight, were changes in population,
reform was “exogenous enough” for the variables
share of females and youths, income per tax tax-
we are interested in. To provide evidence for this,
payer, unemployment share and the share of disa-
we show that the treatment and control groups have
bility pensioners.
common trends in our variables of interest in the period prior to the reform.
40
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
The change in the scheme in 2000 is previously analysed by Stokke
(2016).43
the case we will look at in the following, could lead
In this part of the evalu-
to higher wages since workers will want to partake
ation we will follow Stokke’s approach and focus on
in the firm’s improved profitability. Further, a reduc-
the municipalities that moved from Zone 2 to Zone
tion in labour costs could lead to an increase in em-
3. This group is most suited for several reasons.
ployment, since labour becomes relatively cheaper
First, we need a large amount of data to be confi-
than it previously was and compared to other factors
dent in our results. Second, there is a potential issue
of production, such as capital.
of the prevalence of “commuter
municipalities” 44
among municipalities that moved from Zone 2 to Zone 1. Note that it was the worker’s municipality of residence that mattered for the differentiation of payroll tax in this period. This was changed from worker to firm location in 2007.
4.1.1 Sample construction and restrictions There are several issues to consider in our empirical analysis. We are studying an eight-year period and a policy with many elements, many of which changed during those years. During our estimation
Most workers commuting across tax zones will, in most cases, commute to a zone with a higher tax rate than the tax rate in the zone in which they live.
period (1996-2003), several industry exemptions were put in place in accordance with ESA rulings. Firms in these industries paid the general payroll tax
This is because the payroll tax rate is higher in “well-
rate of 14.1 pct. regardless of geographic location.
performing municipalities”. So, if someone com-
These industries are:
mutes to a municipality where the payroll tax rate is different, it is most likely a well-performing municipality, since it has job opportunities. This means that the worker’s municipality of residence is part of a different labour market, in terms of characteristics, not just geographically, than municipalities that form
▪
Production and distribution of electricity
▪
Extraction of crude petroleum and natural gas
▪
Services activities incidental to oil and gas extraction excluding surveying
▪
nium and thorium ores, as well as some firms in
whole labour markets or economic regions with the same payroll tax rate in all parts of the region. In such (latter) regions, the price of labour is the same in all parts of the region. Thus, comparing or grouping labour markets with differing payroll tax rates to labour markets with nondifferentiated payroll tax rates, could potentially lead to biased estimates.
Mining of non-ferrous metal ores, except uramining of chemical and fertilizer materials
▪
Building and repairing of ships
▪
Manufacturing of basic iron and steel and of ferro-alloys
▪
Financial intermediation
▪
Freight transport by road (firms with more than 50 full-time employees)
We are interested in estimating the effect of a
▪
Telecommunications
change in the payroll tax rate on wages and employment. A reduction in labour costs, which occurs in
In addition to excluding firms in these industries 45, we also exclude the public and primary sectors 46.
43
Currently only available in a working paper version. See http://wwwsre.wu.ac.at/ersa/ersaconfs/ersa16/Paper169_HildegunnStokke.pdf 44 Municipalities where a large part of the workforce is employed in another municipality. In this case, many of the municipalities are neighbours of Bergen.
45
This means dropping 5 pct. of worker-year observations in the treatment and control regions. 46 This means dropping 39 pct. of worker-year observations in the treatment and control regions.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
41
We exclude the primary sector because of the ex-
We exclude individuals and firms with missing infor-
tensive subsidies and the considerable degree of
mation on variables included in the regressions, as
self-employment in these industries. Regarding the
well as individuals and firms who are not present
public sector, the centralised wage bargaining and
both before and after treatment is effectuated or
national regulation with respect to public sector
have “holes” in their time series51. We further restrict
wages warrants their exclusion from our wage re-
our observations to include only those who are ob-
gressions. However, we also exclude them from our
served for at least three consecutive years. In re-
subsequent regressions, to ensure a consistent
gressions at the firm level, we only include firms lo-
data set.
cated in the regions of interest.
We exclude high-paid workers47 and those aged
Figure 4.1 Municipalities in the estimation sample by treatment and control group
above 55 years48 due to other changes in the scheme that could potentially affect our identification of the change in the payroll tax rate (see Chapter 2.3 for information about these changes). We exclude municipalities in economic regions where more than one payroll tax rate is prevalent. Each economic region constitutes one labour market, meaning there is a great deal of commuting between municipalities within the region, but relatively little commuting out of the region. We exclude individuals working in Zone 4 and 549, which is the case for only a very small part of our sample, since these workers travel quite far to find work. We choose to estimate effects at the firm level50, not the establishment level. The data quality is better at the firm level and workers sometimes change employer within the same firm (i.e. change establishment within the same firm). In addition, we avoid potential issues related to re-organisation of establishments within firms, where divisions are split up or
Note: Only economic regions where the payroll tax rate is the same for all municipalities are included in the sample. Source: Samfunnsøkonomisk analyse AS Map: ©Kartverket
grouped together.
47
This means dropping 0.2 pct. of worker-year observations in the treatment and control regions. 48 This means dropping 12 pct. of worker-year observations in the treatment and control regions. 49 This means dropping 0.4 pct. of worker-year observations in the treatment and control regions.
42
50
We use firm to address what Statistics Norway call enterprise in their StatBank. 51 This means dropping 42 pct. of worker-year observations in the treatment and control regions.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
To avoid skewed results due to outliers in our re-
1999, the population dropped marginally more in
gressions, we trim our sample by dropping the top
treatment regions than in control regions. Average
and bottom 2 pct. of the distribution of the depend-
net immigration and unemployment are also equiv-
ent variable. Our results affect our recommenda-
alent between the two groups, as both face net out-
tions regarding a large and important scheme and
migration in the pre-reform period (1996-1999) and
we do not want observations of a small minority to
have an unemployment share of 1.5 pct. In the post-
influence them. We comment on the significance of
reform period, the share of municipalities on the
this trimming in each of the results sections.
ROBEK list was much higher in the control regions, suggesting poor economic conditions and/or finan-
We exclude firms who have establishments outside
cial management in municipalities in this group.
the economic regions that define our treatment and control groups (see Figure 4.1). In addition, we remove firms with employees from both groups. The
Table 4.4 Descriptive statistics on municipality level. Means over the period 1996-1999 Treatment Control group group 23 36
differentiated payroll tax rate lowered labour costs in the regions that are affected. In the period 19962003, this made employees residing in these regions relatively cheaper labour, compared to those residing in the unaffected regions. The objective, then, was for these residents to gain employment or
Variable No. of municipalities1 Payroll tax zone after 19991 Percentage point change in tax
rate1
3
2
-4.2
0
Population
72,989 135,213
Population growth
-0.58 %
-0.06 %
remain employed, not for workers commuting from
Net
- 124
- 206
other tax zones to gain employment.
Unemployment3
1.5 %
1.5 %
Share of mun. on the ROBEK list4
22 %
39 %
To study employment effects, we have aggregated our individual-level data to the firm level. This allows us to study employment effects for those residing in the affected regions, since we can count the number of employees in each firm that resides in any given municipality. Our dependant variable is therefore based on employees residing in the treatment and
immigration2
1) Not means. 2) From other parts of Norway. 3) As share of population aged 15-74 4) Indicates either poor financial management of municipal finances or poor economic conditions. Included if on the ROBEK list at some point during 2001-2004. Note that most were taken off the list one year (or less) after inclusion Sources: Statistics Norway, Samfunnsøkonomisk analyse AS and www.government.no
control regions only, not total employment, which could include commuters from other regions. Con-
Table 4.5 and 4.6 presents descriptive statistics
sequently, we estimate employment growth for
from our individual- and firm-level data sets. Differ-
treated and control regions. This way, we examine
ences between the two tables are largely due to the
the “pure employment effect” for those that were af-
inclusion of part-time workers in the firm level data,
fected by the policy.
which most notably affects the share of female workers, and the fact that the firm level data partly
Table 4.4 present some descriptive statistics for the
includes data based on workers not residing in the
treatment and control regions. The population in the
treatment and control regions. However, data on
control regions is almost twice as large as in the
worker characteristics is based on the employees
treatment regions, although this follows from the
residing in either treated or control regions.
larger number of municipalities. In the period 1996-
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
43
The two groups are comparable at the individual level in terms of the various characteristics, although there is a difference in levels of average
Table 4.6 Descriptive statistics for firm-level data of treatment and control groups. Means over the period 1996-2003
hourly wages. The hourly wage is contracted wage divided by contracted hours and does not include overtime or other forms of payment, but it is internally consistent.
Table 4.5 Descriptive statistics for individual-level data of treatment and control groups. Means over the period 1996-2003 Variable Worker-year observations
Treatment Control group group 21 694
39 712
Hourly wage (2015-NOK)
165
172
Wage growth (2015-NOK)
5.6
5.7
25-34 years old
29.2 %
30.9 %
35-44 years old
36.9 %
36.2 %
45-55 years old
33.9 %
32.9 %
Share of immigrants
1.8 %
3.0 %
19.2 %
22.0 %
Primary education
22.3 %
23.3 %
Secondary education
71.0 %
68.7 %
6.8 %
7.9 %
Age composition
Share of female workers Education composition
Higher education Industry/sector composition Manufacturing
35.1 %
30.2 %
Secondary sector
51.5 %
46.9 %
Tertiary sector
48.5 %
53.1 %
Share of commuters1
8.9 %
7.5 %
Level 4
30.9 %
23.6 %
Level 5
51.0 %
62.1 %
Level 6
18.1 %
14.3 %
Worker-year obs. by centrality2
Notes: Statistics are reported after doing the same adjustments as done prior to estimation. 1) Commuters out of tax zone as share of group. 2) Percentage of group residing in a municipality with the given centrality level. Source: Samfunnsøkonomisk analyse AS
44
Variable Firm-year observations
Treatment Control group group 3 577
8 237
Firm size (reported)
9.6
9.5
Firm size (calculated)1
7.0
6.8
0.4 %
-0.5 %
-0.3 %
3.2 %
25-34 years old
32.9 %
34.9 %
35-44 years old
35.4 %
35.5 %
45-55 years old
31.7 %
29.6 %
Share of immigrants
2.1 %
2.0 %
36.0 %
35.8 %
Primary education
23.4 %
24.9 %
Secondary education
66.7 %
63.3 %
Higher education
10.0 %
11.8 %
Manufacturing
28.7 %
23.6 %
Secondary sector
47.3 %
39.4 %
Tertiary sector
52.7 %
60.6 %
Share of commuters2
2.6 %
2.7 %
Level 4
33.7 %
27.5 %
Level 5
46.7 %
58.6 %
Level 6
19.6 %
14.0 %
76.7 %
78.1 %
Value added growth Capital services growth Age composition
Share of female workers Education composition
Industry/sector composition
Firm-year obs. by centrality3
Share of full-time workers
Notes: Statistics are reported only for firms located in the treatment and control regions. Statistics are reported after doing the same adjustments as done prior to estimation 1) Based on observations in employer-employee register for employees residing in the treatment and control regions. 2) Commuters out of tax zone as share of observed workers in individual level dataset. 3) Percentage of group located in a municipality with the given centrality level. Source: Samfunnsøkonomisk analyse AS
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
At the firm level, the treatment group has more firms
set and use the change in log hourly wages as de-
in manufacturing than the control group, and rela-
pendant variable. Doing this, we account for unob-
tively more employees in this group as well. This
served individual level variation in wages. We esti-
could potentially be a problem in our efforts to iden-
mate the following equation
tify causal effects but will be tested for. Note that no firms in our estimation sample are lo-
∆ ln 𝑤𝑖𝑗𝑠𝑟𝑡 = 𝑎0 + 𝑎1 𝑇𝑟 + 𝑎2 𝑃𝑡 + 𝑎3 𝑇𝑟 𝑃𝑡 + 𝑋̅𝑗𝑡 𝛽 + 𝜑𝑟 + 𝜌𝑡 + 𝜇𝑠 𝜌𝑡
cated outside the treatment and control regions.
+ 𝜀𝑖𝑗𝑠𝑟𝑡
(4.1)
The number of employees residing in the regions as a share of the total number of employees in the firms
where ∆ ln 𝑤𝑖𝑗𝑠𝑟𝑡 is the change in log hourly wage
is about 91 pct.
from year 𝑡 − 1 to year 𝑡 for worker 𝑖 in firm 𝑗 in industry 𝑠 located in region 𝑟, Tr is a dummy that
4.1.2 Econometric strategy Our regression models follow Stokke (2016), though with some modifications. That is, we estimate effects on our selected dependent variables with a difference-in-differences (DiD) approach, using indicator variables to evaluate the effects of a change in the payroll tax rate on various dependent variables, most notably growth in wages and employment. As noted in Stokke (2016), the methodological challenge in the study of effects of the payroll tax rate in Norway is that municipalities are not randomly cho-
equals 1 if the labour market region is part of the treatment group facing lower payroll tax rate, and Pt is a dummy that equals 1 in the post reform years (from 2000 onwards). The vector of worker charac̅ it , includes dummies for age (5teristics in year 𝑡, X year intervals), education level (primary, secondary and collage), immigrant status (native, western immigrant, non-western immigrant) and gender. Regional and year fixed effects are represented by 𝜑𝑟 and 𝜌𝑡 , respectively.52 Industry times year fixed effects capture industry-specific trends and shocks (𝜇𝑠 𝜌𝑡 ). 𝑎0 is a constant, 𝛽 is a vector of parameters and 𝜀𝑖𝑗𝑠𝑟𝑡 is an error term.
sen to have a low or high tax rate. Rather, they are picked based on economic and demographic indicators, meaning there are differences in characteristics between those with high and low tax rates. Stokke (2016) argues that this can be solved by exploiting the payroll tax reform in 2000 and using the municipalities that remained in Zone 2 as controls for those who were moved to Zone 3 with a lower
We are particularly interested in estimating the parameter 𝑎3 , which captures the difference in wage growth between treatment and control regions after the treatment regions change tax zone, compared to the pre-treatment period. We also adjust the above specified regression model to allow year-specific treatment effects.
payroll tax rate. Next, we look at the effect of lower payroll taxes on First, we estimate the effect of a lower payroll tax rate on wages. We apply our individual level data
We use a measure of the municipalities’ centrality. The index is measured according to distance to workplaces and service functions such as retail. Both the treatment group and the control group are represented in each of the centrality levels used. Controlling for municipality or economic 52
employment growth. To do this we use firm-level data, aggregated from our individual-level data and
region would be correlated with treatment, thus creating a problem for identification.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
45
use the change in log firm size as dependent varia-
same percentage change invokes in a firm with 10
ble, measured as the number of workers in the firm.
employees.
Thus, this approach deals with the extensive margin (hired/not hired), as opposed to the intensive margin
The difference-in-differences method hinges on an
(part-time/full-time or number of hours).53 Our re-
assumption of parallel trends, which means that the
gression model is as follows
treatment group would follow the same trend as the control group in absence of treatment. This is called
∆ ln 𝑠𝑖𝑧𝑒𝑗𝑠𝑟𝑡 = 𝑏0 + 𝑏1 𝑇𝑟 + 𝑏2 𝑃𝑡 + 𝑏3 𝑇𝑟 𝑃𝑡 + 𝜑𝑟 + 𝜌𝑡 + 𝜇𝑠 𝜌𝑡 + 𝜀𝑗𝑠𝑟𝑡
(4.2)
a counterfactual, since it is only hypothetical and cannot be observed. We cannot test for it empirically, but we can show some descriptive statistics
where ∆ ln 𝑠𝑖𝑧𝑒𝑗𝑠𝑟𝑡 is the change in log number of
and perform placebo checks54 in our regressions to
workers from year 𝑡 − 1 to year 𝑡 for firm 𝑗 in industry 𝑠 located in region 𝑟, 𝑏0 is a constant and 𝜀𝑗𝑠𝑟𝑡 is an
investigate the pre-treatment trends and group
error term. Other explanatory variables are ex-
ity of the assumption of parallel trends if we find
plained in relation to the wage equation above. Our parameter of interest is 𝑏3 , which captures the
trends that are not parallel prior to treatment.
difference in employment growth between treatment
wages in both treatment and control regions. Nota-
and control regions in the years after the payroll tax
bly, the wage growth rates in the groups are quite
cut, compared to the pre-reform period. We also ad-
similar and declining in the pre-reform period, alt-
just the above specified regression model to allow
hough the average growth over this period is slightly
year-specific treatment effects. Further, the firm
higher in the control group. In 2001 the average
level model will be applied with other dependent
growth rate is higher in the treatment group than in
variables as well, namely the growth rates of num-
the control group. Whether there is a statistically
ber of hours worked by employees residing in the
significant difference will be addressed in the re-
treatment and control regions, capital services and
gressions.
characteristics. We cannot be confident in the valid-
Figure 4.2 show average growth rates in hourly
value added. Figure 4.3 show average growth rates in employees The regressions at firm level will be weighted with
residing in treatment and control regions. Employ-
the level of the dependent variable to adjust for the
ment growth (which we measure as growth in em-
fact that a given percentage change of the depend-
ployees residing in either the control or treatment
ent variable has a different absolute effect on levels
regions) is declining for both groups in the pre-re-
according to the size of the level. For example, a
form period. On average, the two groups have a
percentage change in the number of employees in
similar employment growth in the three-year period
a firm with 100 employees entails a different change
1997-1999; the treatment group’s average is 3.5
in number of employees compared to what the
pct. and the control group’s average is 3.4 pct. Post-
53
two groups differ in terms of the dependant variable regardless of treatment, which invalidates the diff-in-diffs method.
We will also use hours worked as the dependent variable to investigate intensive margin effects. 54 “Placebo” refers to the fact that we check whether there is a treatment effect before treatment occurs. If we find a treatment effect before the actual treatment occurs (i.e. a statistically significant placebo effect), then the
46
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
reform, the average growth rate is higher in employment of workers residing in treated regions.
Figure 4.2 Average annual nominal growth rate of hourly wages in treatment and control groups. 1997-2003
Figure 4.4 Average annual growth rate of hours worked in firms in the treatment and control groups. 1997-2003 0,1
0,1
0,08
0,08
0,06
0,06
0,07
0,07
0,04
0,04
0,06
0,06
0,02
0,02
0,05
0,05
0
0,04
0,04
-0,02
-0,02
0,03
0,03
-0,04
-0,04
0,02
0,02
0,01
0,01
0
0 1997 1998 1999 2000 2001 2002 2003 Controls
Treated
Note: Time series are calculated after making the same adjustments as done prior to estimation (see part 4.1.1). Source: Samfunnsøkonomisk analyse AS
Figure 4.3 Average annual growth rate of employees in firms in the treatment and control groups. 1997-2003
0
-0,06
-0,06 1997 1998 1999 2000 2001 2002 2003 Controls
Treated
Notes: Time series are calculated after making the same adjustments as done prior to estimation (see part 4.1.1). Observations are weighted with the number of hours worked by employees residing in the treatment and control regions. Source: Samfunnsøkonomisk analyse AS
Figure 4.5 Average annual growth rate of value added in firms in the treatment and control groups. 1997-2003 0,12
0,12
0,1
0,1
0,08
0,08
0,08
0,08
0,06
0,06
0,06
0,06
0,04
0,04
0,04
0,04
0,02
0,02
0,02
0,02
0
0
0
0
-0,02
-0,02
-0,02
-0,02
-0,04
-0,04
-0,04
-0,04
-0,06
-0,06
-0,06 1997 1998 1999 2000 2001 2002 2003 Controls
Treated
Notes: Time series are calculated after making the same adjustments as done prior to estimation (see part 4.1.1). Observations are weighted with the number of employees residing in the treatment and control regions. Source: Samfunnsøkonomisk analyse AS
-0,06 1997 1998 1999 2000 2001 2002 2003 Controls
Treated
Notes: Time series are calculated after making the same adjustments as done prior to estimation (see part 4.1.1). Observations are weighted with the level of value added Source: Samfunnsøkonomisk analyse AS
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
47
Using data on the annual number of hours worked per firm, we can estimate the effect of a reduction in the payroll tax rate on employment on the intensive margin. We let this variable define an annual number of hours of labour input in firms, again using our individual level dataset to calculate these hours only for workers residing in control and treatment regions. Figure 4.4 show average growth rates in hours worked by employees residing in treatment and control regions. The historical development is similar to that of the average growth rates of the number of employees.
Figure 4.6 Average annual growth rate of capital services in firms in the treatment and control groups. 1997-2003 0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 -0,05 -0,1 1997 1998 1999 2000 2001 2002 2003
Figure 4.5 show average growth rates in value
Controls
added in firms in the treatment and control regions. The pre-reform trend is similar in the two groups. We test for common trend empirically by running a regression for growth in value added in the period
0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 -0,05 -0,1 Treated
Notes: Time series are calculated after making the same adjustments as done prior to estimation (see part 4.1.1). Observations are weighted by the level of capital services Source: Samfunnsøkonomisk analyse AS
prior to the reform on treatment, trend and an interaction between treatment and trend. The results lead us to dismiss common trend violation.
Figure 4.7 Average annual log of capital services in firms in the treatment and control groups. 19962003
We also investigate the effect of the reduced payroll
5,3
5,3
tax rate on capital. The ideal measure, capturing the economic contribution of capital inputs in a produc-
5,25 5,2
5,2
tion theory context, is flow of capital services (see Draca, Sadun and van Reenen (2007)). Let the variable 𝐾 be a measure of capital services, which are calculated based on the book values of a firm’s tangible assets. Further, all assets are divided in two types: equipment (denoted by the superscript 𝑒) which include machinery, vehicles, tools, and transport equipment; and buildings and land (denoted by the superscript 𝑏). Then capital services 𝐾𝑡 = ∑𝑗=𝑒,𝑏(𝑟 + 𝛿𝑗 )𝐾𝑗𝑡 , where the depreciation rates, 𝛿𝑗 , are 20 pct. for equipment and 5 pct. for buildings (Raknerud, Rønningen and Skjerpen 2007). The real rate of return, 𝑟, is the average real return on
5,15 5,1
5,1 5,05
5
5 4,95
4,9
4,9 1996 1997 1998 1999 2000 2001 2002 2003 Controls
Treated
Notes: Time series are calculated after making the same adjustments as done prior to estimation (see part 4.1.1). Source: Samfunnsøkonomisk analyse AS
10-year government bonds for the period 19992006, and equal to is 4.7 pct. (based on figures from the Norwegian Central Bank).
48
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMSIK-ANALYSE.NO
Figure 4.6 show average growth rates of capital ser-
4.1.3 Results from the wage regressions
vices in the treatment and control groups. We con-
Using the difference-in-differences approach de-
duct the same pre-reform common trend test as we
scribed in the previous section, we estimate the ef-
did with value added above and find that the trend
fects of a reduction in the payroll tax rate on the
is statistically different pre-reform at a 10 per cent
growth in hourly wages. The results of five separate
significance level. If we only include the services
regressions of the impact of the 2000 reform on in-
sector in the test sample, we find that there is a com-
dividual wages for the treatment group is presented
mon trend violation with an even stronger signifi-
in Table 4.7. The parameters of interest are those
cance. Thus, we choose to estimate the treatment
estimated for variables that are interactions be-
effect on the log level of capital services with fixed
tween post reform years and treatment.
effects and a lagged level of capital services. Tests on the level of capital services leads us to accept
Column (1) presents the results from the regression
the hypothesis of common trends, both overall and
restricting the treatment effect to be an average
on the subsamples. Figure 4.7 plots the log level of
over the post reform years, 2000-2003. The point
capital services in the years 1996-2003 for the two
estimate of 0.005 on Treatment x Post 1999 means
groups. The development is similar for the two
that the annual growth rate in hourly wages in the
groups in the period prior to the reform in 2000.
post-reform period is 0.5 percentage points higher in the treatment group than in the control group. The
Some of the growth rate plots bear witness of the
parameter estimate is statistically significant and
business cycles that occurred in our evaluation pe-
relatively robust to model specifications with cluster-
riod. The Norwegian economy left an upturn in 2001
ing at different levels and the log of individual lagged
and entered a downturn, effectuated by the dot-com
wage levels. The implication of this is a cumulative
bubble bursting and worsened by the Norwegian
growth over the post-reform period of 2 pct. The to-
Central Bank’s sharp increase in the interest rate,
tal wage cost reduction on employees in the treat-
which had a large effect on the exchange rates. The
ment group is 3.8 pct.55 Thus, our average result in
latter had an adverse effect for exports and import
Column (1) implies that workers received over half
competition, which meant worse times for manufac-
(53 pct.) of the cost reduction from a reduced payroll
turing. As we saw in Table 4.6, the treatment group
tax rate.
has relatively more activity in manufacturing, which could potentially be a problem for our identification
Column (2) reports the year-specific effects. We see
of a causal effect of the reduction in the payroll tax
that only one post-reform year exhibits a statistically
rate. In the regressions, we include year dummies
significant differing wage growth between the treat-
at industry level which should capture business cy-
ment and control groups, namely 2001. The point
cle effects and common shocks to avoid problems
estimate in 2001 is 0.009. Thus, the difference in
from this composition problem.
wage growth between the treatment and control groups is 0.9 percentage points over the post-reform period, which constitutes about 24 pct. of the
55
Total wage cost is the sum of wage costs and the payroll tax, whereas labour cost is the sum of total wage costs and other personnel costs, where the payroll tax is not applicable. The reduction in the payroll tax rate
of 4.2 percentage points can be shown to correspond to a 3.8 percentage points reduction in total wage cost. The initial level of the tax rate was 10.6 pct. (1.106-1.064)/1.106=0.038.
EVALUATION OF THE REGIONALLY DIFFERENTIATED SOCIAL SECURITY CONTRIBUTIONS IN NORWAY | SAMFUNNSOKONOMISK-ANALYSE.NO
49
reduction in total wage costs following the payroll
between pre-reform years and treatment. The point
tax rate reduction. This is significantly lower than
estimates from Column (2) are only marginally
what the estimate in Column (1) implied. Including a
changed, but the statistical significance of the 2001-
lagged level-version of the dependant variable does
effect does not remain.
not affect the results markedly. Before running the regressions in Table 4.7, we have dropped the top
Column (4) reports the results of a regression using
and bottom two percentiles of observations of the
the same model specification as in Column (2), but
dependant variable. If we exclude another 2 pct. on
only for the subsample of workers employed in the
both sides if the distribution, there are no statistically
secondary sector, dominated by manufacturing.
significant effects when running the same regres-
The point estimates are quite high in 2000 and
sion as in Column (2).
2002, and overall positive. A joint F-test of the sum of the four treatment estimates in Column (4) does
In Column (3) we adopt a more flexible model spec-
not reject the hypothesis that the treatment effect is
ification that controls for potential common trend vi-
positive for the secondary sector and the treatment
olations pre-treatment by including interaction terms
effect is in this case statistically significant at the 5
Table 4.7 Impact of payroll tax cut on individual wage growth
Treatment Post 1999 Treatment x Post 1999
Δ ln w Full sample (1)
Δ ln w Full sample (2)
Δ ln w Full sample (3)
Δ ln w Secondary (4)
Δ ln w Tertiary (5)
-0.002 (0.002) -0.045* (0.019) 0.005** (0.002)
-0.002 (0.002)
0.001 (0.003)
-0.002 (0.002)
-0.002 (0.002)
0.005 (0.004) 0.009** (0.004) 0.004 (0.004) 0.004 (0.003)
0.001 (0.006) -0.003 (0.006) 0.004 (0.006) 0.080 (0.006) 0.003 (0.006) 0.003 (0.006)
0.008 (0.005) 0.006 (0.006) 0.009 (0.006) 0.004 (0.005)
0.001 (0.004) 0.011*** (0.050) -0.001 (0.004) 0.004 (0.005)
Treatment x 1998 Treatment x 1999 Treatment x 2000 Treatment x 2001 Treatment x 2002 Treatment x 2003 Observations
61,406
61,406
61,406
29,765
31,641
Obs. treatment Obs. control
21,694 39,712
21,694 39,712
21,694 39,712
11,159 18,606
10,535 21,106
* p